Online education course evaluation method based on subjective and objective data fusion

Abstract

The invention provides an online education course evaluation method based on subjective and objective data fusion. The method comprises the following steps: constructing an online education course evaluation system; the evaluation system comprises evaluation indexes; obtaining video data and evaluation data of the online education course, and determining evaluation index data according to the video data and the evaluation data; determining a fuzzy evaluation matrix and the weight of the evaluation index according to the evaluation index data; calculating a fuzzy comprehensive matrix according to the weights of the evaluation indexes of the fuzzy evaluation matrix; determining an evaluation result of the online education course according to the fuzzy comprehensive matrix; according to the method, the online education course quality can be accurately and comprehensively evaluated, and the deviation possibly caused by subjective evaluation is reduced.

Description

Technical Field

[0001] The present invention relates to an evaluation field, and in particular to an online education course evaluation method based on the fusion of subjective and objective data. Background Art

[0002] With the rapid development of information technology and the popularization of online education, online education has gradually become an important form of ideological and political education in colleges and universities. With its convenience and flexibility, online education provides a variety of learning methods for college students, allowing education to break through the limitations of time and space and be carried out anytime and anywhere. This emerging form of education not only enriches the means and methods of education, but also provides new possibilities for its contemporary development. However, the development of online education also faces some urgent problems, especially in the evaluation of education quality.

[0003] Existing online education evaluation methods mainly rely on expert evaluation or students' subjective evaluation. This evaluation method has obvious limitations. Expert evaluation is subject to the subjective experience and cognitive bias of experts, and it is difficult to fully and objectively reflect the actual quality of the course. Students' subjective evaluation is often affected by factors such as personal emotions, short-term experience, and preference for teachers, resulting in a certain degree of randomness and uncertainty in the evaluation results. At the same time, due to the lack of effective quality evaluation methods, colleges and universities lack scientific basis when improving course design and teaching methods, which hinders the healthy development and improvement of online education.

[0004] Therefore, in order to solve the above technical problems, it is urgent to propose a new technical means. Summary of the invention

[0005] In view of this, in order to at least solve the problem that the current evaluation method is subject to the subjective experience and cognitive bias of the evaluator and to improve the accuracy of the quality evaluation of online education courses, the present invention proposes an online education course evaluation method based on the fusion of subjective and objective data.

[0006] The present invention provides an online education course evaluation method based on subjective and objective data fusion, comprising the following steps:

[0007] Constructing an online education course evaluation system; the evaluation system includes objective behavior indicators and subjective evaluation indicators;

[0008] Among them, the objective behavior indicators include concentration indicators and positive emotion indicators; the subjective evaluation indicators are satisfaction evaluation indicators; the concentration indicators, positive emotion indicators and satisfaction evaluation indicators are first-level indicators; the concentration indicator includes two second-level indicators, which are the mean of head posture deflection angle and the mean of eyelid closure rate of all users; the positive emotion indicator includes two second-level indicators, which are the proportion of positive expressions and the proportion of positive barrage of all users; the satisfaction evaluation indicators include four second-level indicators, which are the mean of content satisfaction scores, teaching satisfaction scores, interaction satisfaction scores and achievement satisfaction scores of all users;

[0009] Obtaining video data of all users watching online education courses and online education course video data, as well as user evaluation data of online education courses;

[0010] Determine secondary indicator data based on the acquired video data and evaluation data;

[0011] Determine the fuzzy evaluation matrix of the primary indicator according to the secondary indicator data;

[0012] Determine the weights of the primary indicator and the secondary indicator according to the secondary indicator data;

[0013] Calculating a fuzzy comprehensive matrix according to the fuzzy evaluation matrix of the first-level index, the weight of the first-level index and the weight of the second-level index;

[0014] Obtaining membership values ​​of the online education courses at different levels according to the fuzzy comprehensive matrix;

[0015] The evaluation result of the online education course is determined according to the membership value.

[0016] Furthermore, the fuzzy comprehensive matrix is ​​calculated by the following formula:

[0017]

[0018] A i,1 =[ω attention,1 ω attention,2 ]·R i,1

[0019] A i,2 =[ω emotion,1 ω emotion,2 ]·R i,2

[0020] A i,3 =[ω subjective,1 ω subjective,2 ω subjective,3 ω subjective,4 ]·R i,3

[0021] Among them, A i represents the fuzzy comprehensive matrix of the i-th online education course, A i,1 , A i,2 and A i,3 Respectively represent the fuzzy comprehensive matrix corresponding to the concentration index, positive emotion index and satisfaction evaluation index, ω attention ,ω emotion and ω subjective Respectively represent the weights corresponding to the concentration index, positive emotion index and satisfaction evaluation index; ω attention,1 and ω attention,2 They represent the weights corresponding to the mean value of the head posture deflection angle and the mean value of the eyelid closure rate, ω emotion,1 and ω emotion,2 Respectively represent the weights corresponding to the proportion of positive expressions and the proportion of positive comments, ω subjective,1 ,ω subjective,2 ,ω subjective,3 and ω subjective,4 They represent the weights corresponding to the mean scores of content satisfaction, teaching satisfaction, interaction satisfaction and achievement satisfaction, respectively. i,1 , R i,2 and R i,3 The fuzzy evaluation matrices represent the concentration index, positive sentiment index, and satisfaction evaluation index of the i-th online education course respectively.

[0022] Furthermore, the weights of the primary and secondary indicators are determined by the following method:

[0023] Obtain the evaluation scores of the experts on the importance of the first-level indicators and the second-level indicators, and calculate the hierarchical analysis weight ω of the first-level indicators based on the hierarchical analysis method according to the evaluation scores of the experts b-AHP and the hierarchical analysis weight ω of the secondary indicators b,a-AHP ;

[0024] According to the secondary index, the entropy weight ω of the primary index is calculated based on the entropy weight method b-Entropy and the entropy weight ω of the secondary index b,a-Entropy ;

[0025] According to the hierarchical analysis weight ω of the first-level indicators b-AHP , the entropy weight of the first-level indicator ω b-Entropy , the hierarchical analysis weight of the secondary index ω b,a-AHP and the entropy weight ω of the secondary index b,a-Entropy The weights of the first-level indicators and the second-level indicators are calculated according to the following formula:

[0026]

[0027] Among them, ω b Represents the weight of the b-th first-level indicator, b = attention, emotion, subjective, corresponding to the weights of the concentration index, positive emotion index and satisfaction evaluation index, respectively, ω b,a It represents the weight of the ath secondary indicator under the bth primary indicator.

[0028] Furthermore, the fuzzy evaluation matrix R corresponding to the concentration index, positive emotion index and satisfaction evaluation index is i,1 , R i,2 and R i,3 Calculated by the following formula:

[0029]

[0030] Among them, u i,j represents the jth secondary indicator data of the ith online education course, j = 1, 2, ..., 8, corresponding to the mean value of head posture deflection angle, the mean value of eyelid closure rate, the proportion of positive expressions, the proportion of positive comments, the mean value of content satisfaction score, the mean value of teaching satisfaction score, the mean value of interaction satisfaction score and the mean value of achievement satisfaction score, respectively. C n represents the nth level of the rating set, C n (u i,j ) represents the score when the j-th secondary indicator data of the ith online education course is rated as the n-th level.

[0031] Furthermore, the score C of the jth secondary indicator data of the i-th online education course when it is rated as the nth level n (u i,j ) is determined by the following method:

[0032] When the j-th secondary indicator data is positive data, the following formula is used for calculation:

[0033]

[0034]

[0035] When the j-th secondary indicator data is negative, the following formula is used for calculation:

[0036]

[0037] Among them, a i,j,n represents the threshold for the jth secondary indicator data of the ith online education course to be rated as the nth level, u i,jrepresents the j-th secondary indicator data of the ith online education course, j = 1, 2, …, 8, corresponding to the mean head posture deflection angle, the mean eyelid closure rate, the proportion of positive expressions, the proportion of positive barrage, the mean content satisfaction score, the mean teaching satisfaction score, the mean interaction satisfaction score and the mean outcome satisfaction score, respectively.

[0038] Furthermore, the mean value of the head posture deflection angle is determined by video data of the user watching an online education course:

[0039] The video data is sliced, and facial images are extracted from the sliced ​​videos. Based on the head posture estimation algorithm, the pitch angle and yaw angle of the user when watching the online video are obtained according to the extracted facial images, and the mean value of the head posture deflection angle is calculated according to the obtained pitch angle and yaw angle. The calculation formula is as follows:

[0040] attention-head i =max(pitch i ,yaw i )

[0041]

[0042] Among them, attention-head i represents the mean head posture deflection angle of the i-th online education course, pitch i represents the mean pitch angle of the i-th online education course, yaw i represents the mean yaw angle of the i-th online education course, pitch i,d,e,l and yaw i,d,e,l denote the pitch angle and yaw angle of the lth frame facial image extracted from the eth video clip of the dth user of the ith online education course, respectively. D denotes the total number of users of the online education course, E denotes the total number of video clips, and L denotes the total number of frames of extracted facial images.

[0043] Furthermore, the mean eyelid closure rate is determined by video data of users watching online education courses:

[0044] The video data is sliced, and facial images are extracted from the sliced ​​videos. Based on the eye key point recognition algorithm, the eyelid closure rate of the user in different video clips is extracted according to the extracted facial images, and the eyelid closure rate mean is calculated according to the eyelid closure rate of the user in different video clips. The calculation formula is as follows:

[0045]

[0046] Among them, attention-eye irepresents the mean eyelid closure rate of the ith online education course, g represents the eye closure area threshold, Pg i,d,e It represents the percentage of the time when the eye closure rate exceeds g in the e-th video clip of the d-th user of the i-th online education course to the total duration of the video clip.

[0047] Furthermore, the proportion of positive expressions is determined by video data of users watching online education courses:

[0048] The video data is sliced, and facial images are extracted from the sliced ​​video. Based on the facial expression recognition algorithm, the number of frames of the user's positive expression in the i-th online education course is obtained according to the extracted facial images, and the positive expression ratio is calculated according to the number of frames of the user's positive expression in the i-th online education course. The calculation formula is as follows:

[0049]

[0050] Among them, emotion-face i represents the proportion of positive expressions in the i-th online education course, represents the number of positive expression frames of the dth user in the ith online education course, Represents the total number of frames of video data watched by the d-th user of the ith online education course.

[0051] Furthermore, the proportion of positive comments is determined by the following method:

[0052] The online education course video data is obtained, and the bullet screen is divided based on the bullet screen emotion recognition algorithm, and the number of positive bullet screens is determined. The proportion of positive bullet screens is calculated according to the number of positive bullet screens. The calculation formula is as follows:

[0053]

[0054] Among them, emotion-word i represents the proportion of positive comments in the i-th online education course, W i represents the total number of bullet comments for the i-th online education course, Represents the number of positive comments for the i-th online education course.

[0055] Further, the evaluation result of the online education course is an evaluation grade or score;

[0056] The evaluation level is determined by the following method:

[0057] According to the maximum membership principle, the level corresponding to the maximum membership value is determined as the evaluation level of the i-th online education course;

[0058] The score is calculated by the following formula:

[0059] value(A i )=s 1 ×A i (C 1 )+s 2 ×A i (C 2 )+…+s n-1 ×A i (C n-1 )+s n ×A i (C n )

[0060] Among them, value(A i ) represents the score of the ith online education course, s n Indicates the score of the nth level, A i (C n ) represents the degree of membership when the i-th online education course is rated as the n-th level.

[0061] Beneficial effects of the present invention: The present invention evaluates online education courses by collecting objective behavior indicators and subjective evaluation indicators, thereby reducing the deviation that may be caused by subjective evaluation, and can accurately and comprehensively evaluate the quality of online education courses and enhance the credibility of the evaluation; and objective behavior indicators can provide a more comprehensive evaluation perspective, can truly reflect the user's learning status, not only helps to understand the user's learning process, but also can reveal potential learning problems, thereby providing a basis for improving teaching. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0063] Figure 1 It is a schematic diagram of the process of the present invention.

[0064] Figure 2 Schematic diagram of the head posture deflection angle in this embodiment. DETAILED DESCRIPTION

[0065] The present invention is further described below in conjunction with the accompanying drawings:

[0066] The present invention provides an online education course evaluation method based on subjective and objective data fusion, comprising the following steps:

[0067] Constructing an online education course evaluation system; the evaluation system includes objective behavior indicators and subjective evaluation indicators;

[0068] Among them, the objective behavior indicators include concentration indicators and positive emotion indicators; the subjective evaluation indicators are satisfaction evaluation indicators; the concentration indicators, positive emotion indicators and satisfaction evaluation indicators are first-level indicators; the concentration indicator includes two second-level indicators, which are the mean of head posture deflection angle and the mean of eyelid closure rate of all users; the positive emotion indicator includes two second-level indicators, which are the proportion of positive expressions and the proportion of positive barrage of all users; the satisfaction evaluation indicators include four second-level indicators, which are the mean of content satisfaction scores, teaching satisfaction scores, interaction satisfaction scores and achievement satisfaction scores of all users;

[0069] Obtaining video data of all users watching online education courses and online education course video data, as well as user evaluation data of online education courses;

[0070] Determine secondary indicator data based on the acquired video data and evaluation data;

[0071] Determine the fuzzy evaluation matrix of the primary indicator according to the secondary indicator data;

[0072] Determine the weights corresponding to the primary indicator and the secondary indicator according to the secondary indicator data;

[0073] Calculating a fuzzy comprehensive matrix according to the fuzzy evaluation matrix of the first-level index, the weight of the first-level index and the weight of the second-level index;

[0074] The evaluation results of online education courses are determined according to the fuzzy comprehensive matrix. The above method evaluates online education courses by collecting objective behavior indicators and subjective evaluation indicators, and can accurately and comprehensively evaluate the quality of online education courses.

[0075] In this embodiment, an online education course evaluation system is constructed; the evaluation system includes objective behavior indicators and subjective evaluation indicators;

[0076] Among them, the objective behavior indicators include concentration indicators and positive emotion indicators; the subjective evaluation indicators are satisfaction evaluation indicators; the concentration indicators, positive emotion indicators and satisfaction evaluation indicators are first-level indicators; the concentration indicator includes two second-level indicators, namely the mean value of head posture deflection angle and the mean value of eyelid closure rate of all users; the positive emotion indicator includes two second-level indicators, namely the proportion of positive expressions and the proportion of positive barrage of all users; the satisfaction evaluation indicator includes four second-level indicators, namely the mean content satisfaction score, the mean teaching satisfaction score, the mean interaction satisfaction score and the mean achievement satisfaction score of all users. The subjective and objective multi-dimensional evaluation system established not only takes into account the continuous characteristics of the user's online objective behavior, namely concentration and positive emotion; but also takes into account the discrete characteristics of the offline subjective evaluation of course quality, namely satisfaction in content, teaching, interaction and achievements, which can reduce the deviation that may be caused by subjective evaluation and improve the accuracy of online education course quality evaluation.

[0077] In this embodiment, video data of online education courses watched by users and online education course video data, as well as evaluation data of online education courses by users are obtained; secondary indicator data are determined based on the obtained video data and evaluation data;

[0078] The mean value of the head posture deflection angle, the mean value of the eyelid closure rate and the proportion of positive expressions are determined through video data of users watching online education courses:

[0079] First, determine the time window for data analysis, preferably 30 seconds, and slice the video data using the time window. The sliced ​​video data is expressed as:

[0080] video i.d (x, y, z, t) = [video i.d，1 (x,y,z,30),...,video i.d，E (x, y, z, 30)]

[0081] Among them, video i.d (x, y, z, t) represents the sliced ​​video data of the dth user of the ith online education course, video i.d,E (x, y, z, 30) represents the E-th 30-second video clip data of the d-th user of the i-th online education course; x, y and z represent the RGB values ​​of the image, and t represents the video time;

[0082] Since the subsequent analysis is mainly based on video frames, an image extraction of 1 frame per second is performed for each video clip. In this embodiment, a multi-task cascade convolutional neural network MTCNN is used for face recognition cropping, and a three-level network is used to generate a face window and align the face landmark position to extract the RGB picture (224*224) of the face. The extracted facial image is represented as follows:

[0083] Cut(video i.d,e (x,y,z,30))=[face i.d,e，1 (x,y,z),...,face i.d，e，l (x, y, z)], l = 30

[0084] Among them, Cut represents the face recognition cropping function, face i.d,e,l (x, y, z) represents the facial image of the dth user in the eth 30-second video clip of the ith online education course;

[0085] Among them, the process of extracting facial images using a multi-task cascade convolutional neural network is a prior art and will not be described in detail here;

[0086] The mean value of the head posture deflection angle is determined according to the video data:

[0087] The deflection angle of the human head can be divided into three directions: X, Y, and Z, and is represented by three values: pitch, yaw, and roll. Pitch refers to the pitch angle, which corresponds to the rotation around the X-axis coordinate, indicating that the head is raised up or down; yaw refers to the yaw angle, which corresponds to the rotation around the Y-axis coordinate, indicating that the head is turned left or right; roll refers to the rolling angle, which corresponds to the rotation around the Z-axis coordinate, indicating that the head is tilted left or right, such as Figure 2 As shown;

[0088] Based on the head pose estimation algorithm (Head Pose Estimation), the head pose of the user when watching online videos is obtained, including pitch angle, yaw angle and roll angle; wherein, obtaining head pose estimation based on facial images is a prior art, and the specific acquisition method is not described in detail here;

[0089] According to teaching practice experience, the roll angle has a low correlation with the degree of concentration: for example, students often use their hands or arms to support their heads and look at the screen during lectures, but they are still listening to the lectures. Correspondingly, the pitch angle and yaw angle are highly correlated with the degree of concentration. Therefore, the average head posture deflection angle is calculated based on the pitch angle and yaw angle. The calculation formula is as follows:

[0090] attention-head i =max(pitchi ,yaw i )

[0091]

[0092] Among them, attention-head i represents the mean head posture deflection angle of the i-th online education course, pitch i represents the mean pitch angle of the i-th online education course, yaw i represents the mean yaw angle of the i-th online education course, pitch i,d,e,l and yaw i,d,e,l denote the pitch angle and yaw angle of the lth frame of the facial image extracted from the eth video clip of the dth user of the ith online education course, respectively, D denotes the total number of users of the online education course, E denotes the total number of video clips, and L denotes the total number of frames of the extracted facial images;

[0093] By calculating the average of the head posture deflection angles, it is possible to directly reflect the head direction of all users when watching online education courses, thereby reflecting the user's concentration level in the class.

[0094] The eyelid closure rate mean is determined by the video data:

[0095] A large number of studies have shown that people have different behavioral patterns of eye movements under different fatigue states, such as differentiated blinking frequency, rate, line of sight direction, eyelid closure rate, etc. Among them, the percentage of eyelid closure over the pupil over time (PERCLOS) is the most representative eye fatigue indicator. The evaluation of whether the eyes are closed is obtained by analyzing the state of the eyelids. Under the premise of the known iris area estimate, different closure thresholds (60%, 70%, 80%) are set, that is, the percentage of eye closure area is considered to have occurred. For example, P60 represents the total time percentage of the eyes closed for at least 60%. In actual research, P80 is mainly used as an evaluation representation of different fatigue states, that is, when the eye closure area exceeds 80%, it is considered to be closed. In this embodiment, Pg is preferably P80;

[0096] Based on the eye key point recognition algorithm and related tool libraries, such as OpenFace, mediapipe, etc., the key feature points of the eye can be quickly detected, so as to extract P80 of different video clips, and the eyelid closure rate mean is calculated according to the eyelid closure rate of the user in different video clips. The calculation formula is as follows:

[0097]

[0098] Among them, attention-eye i represents the mean eyelid closure rate of the ith online education course, g represents the eye closure area threshold, Pg i,d,e represents the percentage of the time when the eye closure rate of the e-th video clip of the d-th user of the i-th online education course exceeds g to the total duration of the video clip;

[0099] By calculating the mean eyelid closure rate, it is possible to directly reflect the user's fatigue level when watching online education courses, thereby reflecting the user's concentration.

[0100] The user's concentration is comprehensively reflected by the average value of the head posture deflection angle and the average value of the eyelid closure rate. This can avoid the errors caused by data in some special cases and improve the accuracy of concentration evaluation.

[0101] The positive expression ratio is determined by the video data:

[0102] The number of frames of the user's positive expression in the i-th online education course is obtained based on the facial expression recognition algorithm, wherein the facial expression recognition algorithm includes emotion recognition models of infrastructure such as VGG and Resnet (Residual Network); no restriction is made on the specific facial expression recognition algorithm, and the process of using the existing emotion recognition model to recognize positive expressions is a prior art and is not described in detail here;

[0103] The positive expression ratio is calculated according to the number of frames of the user's positive expression in the i-th online education course, and the calculation formula is as follows:

[0104]

[0105] Among them, emotion-face i represents the proportion of positive expressions in the i-th online education course, represents the number of positive expression frames of the dth user in the ith online education course, Represents the total number of frames of video data watched by the d-th user of the ith online education course.

[0106] By calculating the proportion of positive expressions, we can understand the audience's emotional response to the video content, thereby evaluating the popularity of the video and the audience's preferences, which helps video producers adjust content strategies to better meet audience needs.

[0107] The proportion of active bullet comments is determined by the following method:

[0108] As the number of online audio-visual users in my country continues to increase, the barrage culture has developed rapidly, which is a method for users to evaluate the content of videos in real time, and has gradually become an important part of the online video experience. The essence of barrage is text language, so the emotional components of each barrage in different videos can be evaluated with the help of sentiment dictionary analysis, machine learning and deep learning. The process of evaluating the emotional components of barrage is an existing technology and will not be elaborated here;

[0109] The number of positive bullet comments is obtained, and the positive bullet comment ratio is calculated according to the number of positive bullet comments. The calculation formula is as follows:

[0110]

[0111] Among them, emotion-word i represents the proportion of positive comments in the i-th online education course, W i represents the total number of bullet comments for the i-th online education course, Represents the number of positive comments for the i-th online education course.

[0112] By calculating the proportion of positive comments, we can evaluate users' emotions towards online videos from another perspective.

[0113] Comprehensively evaluating users' positive emotions towards online videos through the proportion of positive expressions and the proportion of positive comments can reduce the errors caused by special circumstances. For example, some users like to listen to online education course G, but hate the teacher of online education course G. Therefore, in the process of actively listening to the class, they posted a large number of negative comments. If only a single indicator is used for evaluation, certain errors will occur. Combining the above content, using multiple factors for comprehensive evaluation can not only make a more comprehensive and accurate evaluation, but also reduce the errors caused by a single factor.

[0114] The satisfaction evaluation index is determined by the following method:

[0115] Subjective satisfaction includes: satisfaction with course content, which users can evaluate from aspects such as the degree to which the course content meets expectations and whether the course combination is reasonable; satisfaction with teaching quality, which users can evaluate from aspects such as the teacher's teaching style and expression ability; satisfaction with interactive experience, which users can evaluate from aspects such as the course question-and-answer session and the teacher-student communication session setting; satisfaction with learning outcomes, which users can evaluate from aspects such as whether new skills have been mastered and the feeling of self-improvement; the scoring criteria are set according to needs or experience, such as the SAE10-level scale; based on the obtained subjective satisfaction score, the satisfaction evaluation index is calculated, and the calculation formula is as follows:

[0116]

[0117] Among them, sub-contenti 、sub-quantity i 、sub-interact i and sub-outcome i represents the mean of content satisfaction, teaching satisfaction, interaction satisfaction and achievement satisfaction, d represents the user, D represents the total number of users, sub-content i,d 、sub-quantity i,d 、sub-interact i,d and sub-outcome i,d They respectively represent the content satisfaction score, teaching satisfaction score, interaction satisfaction score and outcome satisfaction score of the d-th user on the ith online education course.

[0118] Satisfaction essentially reflects users' evaluation of the quality of online education courses. It can provide important reference opinions for online education courses, help video producers better understand user needs and improve video quality, and make targeted improvements.

[0119] By comprehensively evaluating online education courses through the above-mentioned subjective and objective indicators, the comprehensiveness and accuracy of the evaluation can be improved, and the one-sidedness caused by a single indicator can be avoided; and obtaining multi-dimensional indicators can help video producers discover deficiencies from multiple angles, thereby optimizing course design and improving course quality.

[0120] In this embodiment, the fuzzy evaluation matrix of the first-level indicator is determined according to the second-level indicator data; the fuzzy evaluation matrix R corresponding to the concentration indicator, positive emotion indicator and satisfaction evaluation indicator is i,1 , R i,2 and R i,3 Calculated by the following formula:

[0121]

[0122]

[0123] Among them, u i,j represents the jth secondary indicator data of the ith online education course, j = 1, 2, ..., 8, corresponding to the mean value of head posture deflection angle, the mean value of eyelid closure rate, the proportion of positive expressions, the proportion of positive comments, the mean value of content satisfaction score, the mean value of teaching satisfaction score, the mean value of interaction satisfaction score and the mean value of achievement satisfaction score, respectively. C n represents the nth level of the rating set, C n (u i,j ) represents the score when the j-th secondary indicator data of the ith online education course is rated as the n-th level.

[0124] The score C when the j-th secondary indicator data of the i-th online education course is rated as the n-th level n (u i,j ) is determined by the following method:

[0125] When the j-th secondary indicator data is positive data, the positive data includes the proportion of positive expressions, the proportion of positive barrage, the mean content satisfaction score, the mean teaching satisfaction score, the mean interaction satisfaction score and the mean achievement satisfaction score; the following formula is used for calculation:

[0126] When the j-th secondary indicator data is positive data, the following formula is used for calculation:

[0127]

[0128]

[0129] When the j-th secondary indicator data is negative, the following formula is used for calculation:

[0130]

[0131] Among them, a i,j,n represents the threshold for the jth secondary indicator data of the ith online education course to be rated as the nth level, u i,j Represents the j-th secondary indicator data of the ith online education course.

[0132] Among them, n is set according to demand, and preferably four levels are set, which are divided into excellent, good, qualified and unqualified; the thresholds of different secondary indicators at different levels are determined according to experience or demand;

[0133] By calculating the fuzzy evaluation matrix of concentration index, positive emotion index and satisfaction evaluation index, the degree of affiliation of each indicator under different evaluation levels can be reflected, providing a data basis for subsequent comprehensive evaluation.

[0134] In this embodiment, the weights corresponding to the primary index and the secondary index are determined according to the secondary index data; the weights of the primary index and the secondary index are determined by the following method:

[0135] Obtain the evaluation scores of the experts on the importance of the first-level indicators and the second-level indicators, and calculate the hierarchical analysis weight ω of the first-level indicators based on the hierarchical analysis method according to the evaluation scores of the experts b-AHP and the hierarchical analysis weight ω of the secondary indicators b，a-AHP ;

[0136] According to the secondary index, the entropy weight ω of the primary index is calculated based on the entropy weight method b-Entropyand the entropy weight ω of the secondary index b，a-Entropy ;

[0137] Among them, the process of calculating weights according to the hierarchical analysis method and the entropy weight method is an existing technology, and its calculation process will not be described in detail here;

[0138] According to the hierarchical analysis weight ω of the first-level indicators b-AHP , the entropy weight of the first-level indicator ω b-Entropy , the hierarchical analysis weight of the secondary index ω b，a-AHP and the entropy weight ω of the secondary index b，a-Entropy The weights of the first-level indicators and the second-level indicators are calculated according to the following formula:

[0139]

[0140] Among them, ω b Represents the weight of the b-th first-level indicator, b = attention, emotion, subjective, corresponding to the weights of the concentration index, positive emotion index and satisfaction evaluation index, respectively, ω b，a It represents the weight of the ath secondary indicator under the bth primary indicator.

[0141] The analytic hierarchy process mainly relies on the experience and judgment of experts to determine the weights, while the entropy weight law determines the weights based on the objective differences in the data. Combining the two not only takes into account the experience and judgment of experts, but also combines the actual differences in the data, which can effectively eliminate the shortcomings of a single method, improve the accuracy of the weights, and make the determination of the weights more scientific and reasonable.

[0142] In this embodiment, a fuzzy comprehensive matrix is ​​calculated based on the fuzzy evaluation matrix of the first-level index, the weight of the first-level index and the weight of the second-level index;

[0143] The fuzzy comprehensive matrix is ​​calculated by the following formula:

[0144]

[0145] A i,1 =[ω attention，1 ω attention，2 ]·R i，1

[0146] A i,2 =[ω emotion,1 ω emotion，2 ]·R i，2

[0147] A i,3 =[ω subjective,1 ω subjective,2 ω subjective,3 ωsubjective,4 ]·R i，3

[0148] Among them, A i represents the fuzzy comprehensive matrix of the i-th online education course, A i，1 , A i，2 and A i，3 Respectively represent the fuzzy comprehensive matrix corresponding to the concentration index, positive emotion index and satisfaction evaluation index, ω attention ,ω emotion and ω subjective Respectively represent the weights corresponding to the concentration index, positive emotion index and satisfaction evaluation index; ω attention，1 and ω attention，2 They represent the weights corresponding to the mean value of the head posture deflection angle and the mean value of the eyelid closure rate, ω emotion，1 and ω emotion，2 Respectively represent the weights corresponding to the proportion of positive expressions and the proportion of positive comments, ω subjective,1 ,ω subjective,2 ,ω subjective,3 and ω subjective,4 They represent the weights corresponding to the mean scores of content satisfaction, teaching satisfaction, interaction satisfaction and achievement satisfaction, respectively. i，1 , R i，2 and R i，3 The fuzzy evaluation matrices represent the concentration index, positive sentiment index, and satisfaction evaluation index of the i-th online education course respectively.

[0149] Fuzzy comprehensive evaluation can effectively deal with the uncertainty in the evaluation process. Through the method of fuzzy mathematics, fuzzy concepts can be converted into quantifiable values, thereby improving the accuracy and reliability of the evaluation.

[0150] In this embodiment, the membership values ​​of the online education courses at different levels are determined according to the fuzzy comprehensive matrix;

[0151] Determining an evaluation result according to the membership value;

[0152] The evaluation result of the online education course is an evaluation grade or score;

[0153] The evaluation level is determined by the following method:

[0154] According to the maximum membership principle, the level corresponding to the maximum membership value is determined as the evaluation level of the i-th online education course.

[0155] The evaluation result of the online education course may also be a score;

[0156] The score is calculated by the following formula:

[0157] value(A i )=s 1 ×A i (C 1 )+s 2 ×A i (C 2 )+…+s n-1 ×A i (C n-1 )+s n ×A i (C n )

[0158] Among them, value(A i ) represents the score of the ith online education course, s n Indicates the score of the nth level, A i (C n ) represents the degree of membership when the i-th online education course is rated as the n-th level.

[0159] The above method avoids the one-sidedness of a single evaluation by combining subjective data with objective data, making the evaluation of online education courses more scientific and reasonable.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. An online education course evaluation method based on subjective and objective data fusion, characterized by: The following steps are involved: Constructing an online education course evaluation system; the evaluation system includes objective behavior indicators and subjective evaluation indicators; Among them, the objective behavior indicators include concentration indicators and positive emotion indicators; the subjective evaluation indicators are satisfaction evaluation indicators; the concentration indicators, positive emotion indicators and satisfaction evaluation indicators are first-level indicators; the concentration indicator includes two second-level indicators, which are the mean of head posture deflection angle and the mean of eyelid closure rate of all users; the positive emotion indicator includes two second-level indicators, which are the proportion of positive expressions and the proportion of positive barrage of all users; the satisfaction evaluation indicators include four second-level indicators, which are the mean of content satisfaction scores, teaching satisfaction scores, interaction satisfaction scores and achievement satisfaction scores of all users; Obtaining video data of all users watching online education courses and online education course video data, as well as user evaluation data of online education courses; Determine secondary indicator data based on the acquired video data and evaluation data; Determine the fuzzy evaluation matrix of the primary indicator according to the secondary indicator data; Determine the weights of the primary indicator and the secondary indicator according to the secondary indicator data; Calculating a fuzzy comprehensive matrix according to the fuzzy evaluation matrix of the first-level index, the weight of the first-level index and the weight of the second-level index; Obtaining membership values ​​of the online education courses at different levels according to the fuzzy comprehensive matrix; An evaluation result of the online education course is determined according to the membership value.

2. The online education course evaluation method based on subjective and objective data fusion according to claim 1 is characterized by: The fuzzy comprehensive matrix is ​​calculated by the following formula: A i,1 =[ω attention,1 oh attention,2 ]·R i,1 A i,2 =[ω emotion,1 oh emotion,2 ]·R i,2 A i,3 =[ω subjective,1 oh subjective,2 oh subjective,3 oh subjective,4 ]·R i,3 Among them, A i represents the fuzzy comprehensive matrix of the i-th online education course, A i,1 , A i,2 and A i,3 Respectively represent the fuzzy comprehensive matrix corresponding to the concentration index, positive emotion index and satisfaction evaluation index, ω attention ,ω emotion and ω subjective Respectively represent the weights corresponding to the concentration index, positive emotion index and satisfaction evaluation index; ω attention,1 and ω attention,2 They represent the weights corresponding to the mean value of the head posture deflection angle and the mean value of the eyelid closure rate, ω emotion,1 and ω emotion,2 Respectively represent the weights corresponding to the proportion of positive expressions and the proportion of positive comments, ω subjective,1 ,ω subjective,2 ,ω subjective,3 and ω subjective,4 They represent the weights corresponding to the mean scores of content satisfaction, teaching satisfaction, interaction satisfaction and achievement satisfaction, respectively. i,1 , R i,2 and R i,3 The fuzzy evaluation matrices represent the concentration index, positive sentiment index, and satisfaction evaluation index of the i-th online education course respectively.

3. The online education course evaluation method based on subjective and objective data fusion according to claim 1 is characterized by: The weights of the first-level indicators and the second-level indicators are determined by the following method: Obtain the evaluation scores of the experts on the importance of the first-level indicators and the second-level indicators, and calculate the hierarchical analysis weight ω of the first-level indicators based on the hierarchical analysis method according to the evaluation scores of the experts b-AHP and the hierarchical analysis weight ω of the secondary indicators b,a-AHP ; According to the secondary index, the entropy weight ω of the primary index is calculated based on the entropy weight method b-Entropy and the entropy weight ω of the secondary index b,a-Entropy ; According to the hierarchical analysis weight ω of the first-level indicators b-AHP , the entropy weight of the first-level indicator ω b-Entropy , the hierarchical analysis weight of the secondary index ω b,a-AHP and the entropy weight ω of the secondary index b,a-Entropy The weights of the first-level indicators and the second-level indicators are calculated according to the following formula: Among them, ω b Represents the weight of the b-th first-level indicator, b = attention, emotion, subjective, corresponding to the weights of the concentration index, positive emotion index and satisfaction evaluation index, respectively, ω b,a It represents the weight of the ath secondary indicator under the bth primary indicator.

4. The online education course evaluation method based on subjective and objective data fusion according to claim 1 is characterized by: The fuzzy evaluation matrix R corresponding to the concentration index, positive emotion index and satisfaction evaluation index i,1 , R i,2 and R i,3 Calculated by the following formula: Among them, u i,j represents the jth secondary indicator data of the ith online education course, j = 1, 2, ..., 8, corresponding to the mean value of head posture deflection angle, the mean value of eyelid closure rate, the proportion of positive expressions, the proportion of positive comments, the mean value of content satisfaction score, the mean value of teaching satisfaction score, the mean value of interaction satisfaction score and the mean value of achievement satisfaction score, respectively. C n represents the nth level of the rating set, C n (u i,j ) represents the score when the j-th secondary indicator data of the ith online education course is rated as the n-th level.

5. The online education course evaluation method based on subjective and objective data fusion according to claim 4 is characterized by: The score C when the j-th secondary indicator data of the i-th online education course is rated as the n-th level n (u i,j ) is determined by the following method: When the j-th secondary indicator data is positive data, the following formula is used for calculation: When the j-th secondary indicator data is negative, the following formula is used for calculation: Among them, a i,j,n represents the threshold for the j-th secondary indicator data of the i-th online education course to be rated as the n-th level, u i,j represents the j-th secondary indicator data of the ith online education course, j = 1, 2, …, 8, corresponding to the mean head posture deflection angle, the mean eyelid closure rate, the proportion of positive expressions, the proportion of positive barrage, the mean content satisfaction score, the mean teaching satisfaction score, the mean interaction satisfaction score and the mean outcome satisfaction score, respectively.

6. The online education course evaluation method based on subjective and objective data fusion according to claim 5 is characterized by: The mean value of the head posture deflection angle is determined by the video data of the user watching the online education course: The video data is sliced, and facial images are extracted from the sliced ​​videos. Based on the head posture estimation algorithm, the pitch angle and yaw angle of the user when watching the online video are obtained according to the extracted facial images, and the mean value of the head posture deflection angle is calculated according to the obtained pitch angle and yaw angle. The calculation formula is as follows: attention-head i =max(pitch i ,yaw i ) Among them, attention-head i represents the mean head posture deflection angle of the i-th online education course, pitch i represents the mean pitch angle of the i-th online education course, yaw i represents the mean yaw angle of the i-th online education course, pitch i,d,e,l and yaw i,d,e,l denote the pitch angle and yaw angle of the lth frame facial image extracted from the eth video clip of the dth user of the ith online education course, respectively. D denotes the total number of users of the online education course, E denotes the total number of video clips, and L denotes the total number of frames of extracted facial images.

7. The online education course evaluation method based on subjective and objective data fusion according to claim 5 is characterized by: The mean eyelid closure rate is determined by video data of users watching online education courses: The video data is sliced, and facial images are extracted from the sliced ​​videos. Based on the eye key point recognition algorithm, the eyelid closure rate of the user in different video clips is extracted according to the extracted facial images, and the eyelid closure rate mean is calculated according to the eyelid closure rate of the user in different video clips. The calculation formula is as follows: Among them, attention-eye i represents the mean eyelid closure rate of the ith online education course, g represents the eye closure area threshold, Pg i,d,e It represents the percentage of the time when the eye closure rate exceeds g in the e-th video clip of the d-th user of the i-th online education course to the total duration of the video clip.

8. The online education course evaluation method based on subjective and objective data fusion according to claim 5 is characterized by: The proportion of positive expressions is determined by video data of users watching online education courses: The video data is sliced, and facial images are extracted from the sliced ​​video. Based on the facial expression recognition algorithm, the number of frames of the user's positive expression in the i-th online education course is obtained according to the extracted facial images, and the positive expression ratio is calculated according to the number of frames of the user's positive expression in the i-th online education course. The calculation formula is as follows: Among them, emotion-face i represents the proportion of positive expressions in the i-th online education course, represents the number of positive expression frames of the dth user in the ith online education course, Represents the total number of frames of video data watched by the d-th user of the ith online education course.

9. The online education course evaluation method based on subjective and objective data fusion according to claim 5 is characterized by: The proportion of active bullet comments is determined by the following method: The online education course video data is obtained, and the bullet screen is divided based on the bullet screen emotion recognition algorithm, and the number of positive bullet screens is determined. The proportion of positive bullet screens is calculated according to the number of positive bullet screens. The calculation formula is as follows: Among them, emotion-word i represents the proportion of positive comments in the i-th online education course, W i represents the total number of bullet comments for the i-th online education course, Represents the number of positive comments for the i-th online education course.

10. The online education course evaluation method based on subjective and objective data fusion according to claim 2 is characterized by: The evaluation result of the online education course is an evaluation grade or score; The evaluation level is determined by the following method: According to the maximum membership principle, the level corresponding to the maximum membership value is determined as the evaluation level of the i-th online education course; The score is calculated by the following formula: value(A i )=s1×A i (C1)+s2×A i (C2)+…+s n-1 ×A i (C n-1 )+s n ×A i (C n ) Among them, value(A i ) represents the score of the ith online education course, s n Indicates the score of the nth level, A i (C n ) represents the degree of membership when the i-th online education course is rated as the n-th level.

Images