Evaluation method for teaching quality of thought classroom

Abstract

The invention relates to the technical field of teaching quality evaluation, and discloses a thought classroom teaching quality evaluation method, which comprises the following steps: step 1, sound decibel detection; step 2, student state detection; step 3, sound data processing; step 4, student state analysis; 5, comprehensively evaluating the model; and step 6, dynamic adjustment. According to the invention, image recognition is carried out by applying a deep learning algorithm, states of students such as class attending, discussion and sleeping can be accurately judged, postures and facial expressions of the students are analyzed through a trained deep learning model, the accuracy of state classification is improved, and the image quality is improved through an image enhancement technology. And the accuracy of image data is ensured by using a camera calibration technology, and the comprehensive evaluation model is ensured to continuously reflect the real teaching quality condition by regularly adjusting the evaluation standard and the feature weight, so that the beneficial effect that factors with insufficient accuracy cannot occur when the teaching quality is detected is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of teaching quality assessment, in particular to a method for evaluating the quality of ideological classroom teaching. Background Art

[0002] Ideological and political courses are social practice activities in which a society or social group uses certain political viewpoints and moral norms to exert purposeful, planned and organized influence on its members, so that they can form ideological and moral qualities that meet the requirements of a certain society. Because of its certain agitational nature, it requires teachers and students to fully participate in physical and psychological aspects. During the class, the teacher's teaching voice, the students' discussion voice, the heart rate of teachers and students and other physical signs will change accordingly. By real-time detection of relevant information of teachers and students, the teaching quality can be evaluated more objectively.

[0003] For example, the technical solution with publication number CN 106780220 B discloses a method for evaluating the teaching quality of ideological and political courses. By real-time detection of the teacher's teaching voice, the student's discussion voice, and the heart rate and physical signs of the teacher and the student, a more objective evaluation of the teaching quality is carried out from the dimensions of the teacher's involvement and the student's participation.

[0004] The above technical solution uses multiple detections to calculate the teaching quality, especially the volume detection module is used to detect the decibel values ​​of teaching sounds and discussion sounds, but the explanation sounds and student discussion sounds will confuse the volume detection module and make it difficult to perform effective measurements. In addition, the noise in the teacher's surroundings also affects the accuracy of the decibel value, disrupting the judgment of the volume detection module. Student status detection uses an image recognition module to judge the student's status of listening to lectures, discussing, and sleeping. Although it helps to understand students' participation, image recognition technology is affected by factors such as lighting and angle, and may deviate. Summary of the invention

[0005] Technical issues solved:

[0006] In view of the deficiencies of the prior art, the present invention provides a method for evaluating the quality of ideological classroom teaching, which has the advantages of no factors of insufficient accuracy when detecting teaching quality and no deviation in the evaluation of teaching quality data, thereby solving the problems of the above-mentioned technology.

[0007] Technical solution:

[0008] To achieve the above object, the present invention provides the following technical solution: a method for evaluating the quality of ideological classroom teaching, comprising the following steps:

[0009] Step 1: Sound decibel detection: Use a microphone array to place multiple microphones at different locations in the classroom to separate and capture sounds from different sources, and use sound source localization and acoustic model analysis to reduce noise interference;

[0010] Step 2: Student status detection: In a multi-camera system, deep learning algorithms are used for image recognition to accurately determine the student status: listening to a class, discussing, or sleeping. The trained deep learning model is used to analyze the student's posture and facial expression, accurately classify the student's status, and combine real-time lighting data and camera angle calibration to further improve the accuracy of status recognition.

[0011] Step 3: Sound data processing: integrate the sound data collected by multiple microphones, apply filters and noise suppression technology to reduce the impact of environmental noise on the sound decibel value, and classify the teaching sound and discussion sound: calculate the decibel values ​​of the teaching sound and discussion sound respectively, and compare them with the reference values ​​in the storage module to obtain the sound-related teaching quality indicators;

[0012] Step 4: Student status analysis: By processing students’ posture images in real time, applying image enhancement technology and image recognition algorithms, we can determine whether students are listening to a class, discussing, or sleeping. We can compare students’ status data with reference values, and combine image recognition results and sound data to evaluate students’ participation.

[0013] Step 5: Comprehensive evaluation model: Combine the sound decibel value and student status information to calculate the teaching quality using a comprehensive evaluation model;

[0014] Step 6: Dynamic adjustment: Adjust the evaluation criteria and algorithms based on the real-time collected data and evaluation results to ensure the accuracy and adaptability of the system;

[0015] The original image in step 4 is I original , the image enhancement function is f, and the image after enhancement is I enhanced , the image enhancement expression is:

[0016] I enhanced =f(I original ,taansformations)

[0017] Where: f(·) is the image enhancement function, which is used to process the original image and apply the specified transformation to generate the enhanced image I enhanced ;

[0018] Suppose the student status prediction after CNN processing is in The category with the highest probability is:

[0019]

[0020] f CNN (I enhanced ) is the enhanced image I through the convolutional neural networkenhanced Process and output a vector representing the probability of each category;

[0021] The expression for comparing the state data with the reference value in step 4 is:

[0022]

[0023] Where: N is the total number of samples; The true label of the i-th sample; Accuracy is the accuracy, which calculates the proportion of all sample predictions that are correct, that is, the proportion of correctly predicted samples to the total number of samples;

[0024] in is the Kronecker delta function:

[0025]

[0026] Preferably, the expression combining the image recognition result and the sound data in step 4 is:

[0027] The participation rate of sound data is P sound , the participation of image recognition is P image , comprehensive participation P total Calculated by weighted average:

[0028] P total =α·P sound +(1-α)·P image

[0029] Where: P sound is the participation of the sound data, indicating the participation degree or weight of the sound analysis results; α is the weight coefficient of the sound data, ranging from 0 to 1, which determines the influence ratio of the sound data in the calculation of the comprehensive participation; P total To synthesize engagement, the impact of sound data and image data was combined;

[0030] Set the threshold as threshold, according to the comprehensive participation P total And threshold for final evaluation:

[0031]

[0032] Where: P total : Comprehensive participation, a comprehensive evaluation result combining sound and image data; S is the final evaluation result, based on P total Whether it is greater than or equal to the threshold determines the level of participation. If P total If it is greater than or equal to the threshold, it is evaluated as "high engagement"; otherwise, it is evaluated as "low engagement".

[0033] Preferably, the sound decibel detection expression in step 1 is:

[0034]

[0035] Where: L is the decibel value of the sound, indicating the sound pressure level; I0 is the reference sound intensity, which is taken as (1×10 -12 )W / m 2 , which is the minimum sound intensity that can be heard by the human ear.

[0036] Preferably, the sound source localization and acoustic model analysis expression in step 1 is:

[0037] Use a microphone array to locate multiple sound sources and calculate the direction θ and position (x, y, z) of the sound source, expressed as:

[0038]

[0039] in: Indicates the location of the sound source in three-dimensional space;

[0040] The acoustic model M is used for analysis and calculation of the sound propagation characteristics, including reflection and refraction phenomena, which can be expressed mathematically as:

[0041]

[0042] Where t is the time variable;

[0043] Reduce noise disturbance:

[0044] The noise interference N is reduced by signal processing techniques and is defined as:

[0045] S clean =S raw –N

[0046] Where: S clean To remove the noise from the sound signal, S raw is the original sound signal.

[0047] Preferably, in step 2: the deep learning model is used for image classification to determine the status of the student, and the expression is:

[0048] state classification = argmax(f(I; Θ))

[0049] Where: I is the input image or video frame; Θ is the parameter of the deep learning model, including the weights of the convolutional layer and the fully connected layer; f is the output function of the deep learning model, which returns the probability distribution of each state; argmax means selecting the state with the highest probability as the classification result;

[0050] Suppose the output of the model is a vector y containing the probability of each state:

[0051] y=[p 听课 ,p 讨论 ,p 睡觉 ]

[0052] Then: state classification = argmax(p 听课 ,p 讨论 ,p 睡觉 ).

[0053] Preferably, the calibration model expression in step 2 is:

[0054] I 校准 =I·K(C)

[0055] in:

[0056] I 校准 is the calibrated image; I is the original image; K(C) is the camera calibration matrix function, which represents the transformation based on the camera configuration information: angle, focal length and position.

[0057] Preferably, in step 3, the sound data fusion: the sound data collected from multiple microphones needs to be fused to obtain a more accurate sound signal. i (t) represents the sound signal collected by the i-th microphone at time t, and the fused sound signal x(t) is expressed as:

[0058]

[0059] Where: N is the number of microphones; x i (t) is the signal collected by the i-th microphone at time t.

[0060] Preferably, the filter and noise suppression expressions in step 3 are:

[0061] Assume that a linear filter is applied, and its filter coefficient is h(t). The signal after filtering is It is expressed as:

[0062]

[0063] Where: * represents the convolution operation, and the filter coefficient h(t) is designed as a low-pass filter or a band-pass filter;

[0064] Noise suppression techniques include noise estimation and noise suppression filters, using spectral subtraction to reduce noise:

[0065]

[0066] in: is the estimated noise signal;

[0067] Calculate the decibel value of the processed signal. Suppose the processed signal is The calculation formula of its decibel value L(t) is:

[0068]

[0069] Where: T is the time window length for calculating decibel values; It is the square value of the signal and represents the power of the signal.

[0070] Preferably, the feature extraction expression in step 5 is:

[0071] The sound decibel value obtained from step 3 is L(t), and the student status classification result obtained from step 2 is S. The features include: sound decibel value: L(t); student status: S: S = listening to the class, S = discussing, or S = sleeping;

[0072] Feature normalization: Standardize different features so that they can be integrated on the same scale. Let the maximum and minimum values ​​of L(t) be L max and L min , then the normalized decibel value L(t) is expressed as:

[0073]

[0074] The student status is converted into numerical features, including: listening to the class: 1; discussing: 2; sleeping: 0;

[0075] Then the status information S norm Take these values ​​directly;

[0076] The normalized features are input into the comprehensive evaluation model, and the weighted sum model is used to calculate the teaching quality index. The comprehensive evaluation Q is expressed as:

[0077] Q=w1·L norm (t)+w2·S norm

[0078] Among them: w1 and w2 are the weights of the features, indicating the contribution of each feature to the teaching quality; the setting of weights w1 and w2 is based on historical data or expert opinions, and the evaluation results are optimized through training models or adjustments.

[0079] Preferably, the evaluation criteria and algorithm expressions are adjusted in step 6 as follows:

[0080] Adjust threshold: Q threshold; Adjust weights: adjust feature weights w1 and w2 according to feature importance, and use regression analysis, optimization algorithms, or expert advice to determine new weights; assuming the new weights are w′1 and w′2, the updated comprehensive evaluation model is:

[0081] Q t =w′1·L norm (t)+w′2·S norm .

[0082] Compared with the prior art, the present invention provides a method for evaluating the quality of ideological classroom teaching, which has the following beneficial effects:

[0083] 1. The present invention can accurately judge the status of students: listening to lectures, discussing, sleeping, etc. by applying deep learning algorithms for image recognition. The posture and facial expression of students are analyzed by trained deep learning models to improve the accuracy of status classification. Image enhancement technology is used to improve image quality, and camera calibration technology is used to ensure the accuracy of image data, thereby improving the accuracy of status recognition. The participation of sound data and the participation of image recognition are weighted averaged, and the influence of sound and image data on teaching quality is comprehensively considered to reduce the error of a single data source. The evaluation criteria and algorithms are dynamically adjusted according to the real-time collected data and evaluation results to ensure that the system can adapt to different teaching environments and conditions, further improve the accuracy, extract features from the sound decibel value and the student status classification results, and perform normalization processing so that each feature is comprehensively evaluated on the same scale to reduce the influence of differences between features on the evaluation results. By regularly adjusting the evaluation criteria and feature weights, it is ensured that the comprehensive evaluation model can continuously reflect the actual teaching quality situation, and the beneficial effect of not having factors of insufficient accuracy when detecting teaching quality is achieved.

[0084] 2. The present invention improves image quality by using image enhancement technology, enhances the accuracy of state recognition, reduces the deviation caused by image quality problems, corrects image distortion by camera calibration technology, ensures the accuracy of image data, reduces the deviation caused by improper camera configuration, and uses a trained deep learning model to classify student states. The model improves prediction accuracy by training with a large number of samples, reduces human factors or model deviations, calculates accuracy by comparing with reference values, ensures accuracy and reliability of state recognition, combines sound data and image data, calculates comprehensive participation by weighted averaging, balances the influence of different data sources, reduces the deviation of a single data source, adjusts evaluation standards and algorithms according to real-time data and evaluation results, adapts to different teaching environments, ensures that the evaluation model can accurately reflect the teaching quality, reduces the deviation caused by fixed standards, and achieves the beneficial effect of no deviation in the evaluation of teaching quality data. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 It is a schematic diagram of the system of the present invention; DETAILED DESCRIPTION

[0086] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0087] See also Figure 1 ,The evaluation method of ideological classroom teaching quality includes the following steps:

[0088] Step 1: Sound decibel detection: Use a microphone array to place multiple microphones at different locations in the classroom to separate and capture sounds from different sources, and use sound source localization and acoustic model analysis to reduce noise interference;

[0089] Step 2: Student status detection: In a multi-camera system, deep learning algorithms are used for image recognition to accurately determine the student status: listening to a class, discussing, or sleeping. The trained deep learning model is used to analyze the student's posture and facial expression, accurately classify the student's status, and combine real-time lighting data and camera angle calibration to further improve the accuracy of status recognition.

[0090] Step 3: Sound data processing: integrate the sound data collected by multiple microphones, apply filters and noise suppression technology to reduce the impact of environmental noise on the sound decibel value, and classify the teaching sound and discussion sound: calculate the decibel values ​​of the teaching sound and discussion sound respectively, and compare them with the reference values ​​in the storage module to obtain the sound-related teaching quality indicators;

[0091] Step 4: Student status analysis: By processing students’ posture images in real time, applying image enhancement technology and image recognition algorithms, we can determine whether students are listening to a class, discussing, or sleeping. We can compare students’ status data with reference values, and combine image recognition results and sound data to evaluate students’ participation.

[0092] Step 5: Comprehensive evaluation model: Combine the sound decibel value and student status information to calculate the teaching quality using a comprehensive evaluation model;

[0093] Step 6: Dynamic adjustment: Adjust the evaluation criteria and algorithms based on the real-time collected data and evaluation results to ensure the accuracy and adaptability of the system;

[0094] The original image in step 4 is I original, the image enhancement function is f, and the image after enhancement is I enhanced , the image enhancement expression is:

[0095] I enhanced =f(I original ,transformations)

[0096] Where: f(·) is the image enhancement function, which is used to process the original image and apply the specified transformation to generate the enhanced image I enhanced ;

[0097] Suppose the student status prediction after CNN processing is in The category with the highest probability is:

[0098]

[0099] f CNN (I enhanced ) is the enhanced image I through the convolutional neural network enhanced Process and output a vector representing the probability of each category;

[0100] The expression for comparing the state data with the reference value in step 4 is:

[0101]

[0102] Where: N is the total number of samples; The true label of the i-th sample; Accuracy is the accuracy, which calculates the proportion of all sample predictions that are correct, that is, the proportion of correctly predicted samples to the total number of samples;

[0103] in is the Kronecker delta function:

[0104]

[0105] By placing multiple microphones at different locations in the classroom, the impact of environmental noise on sound data can be reduced and the accuracy of sound decibel detection can be improved.

[0106] Sound source localization technology is used to determine the source of sound, and the sound propagation characteristics are analyzed through acoustic models, thereby reducing noise interference and improving the accuracy of sound decibel values.

[0107] Use filters and noise suppression techniques (such as spectral subtraction) to further clean up the noise in the sound signal to ensure that the resulting sound data is more accurate.

[0108] Deep learning algorithm: Applying deep learning algorithm for image recognition can accurately judge the status of students (listening to class, discussing, sleeping). Through the trained deep learning model, students’ posture and facial expressions are analyzed to improve the accuracy of status classification.

[0109] Image enhancement technology is used to improve image quality, and camera calibration technology is used to ensure the accuracy of image data, thereby improving the accuracy of state recognition.

[0110] Comprehensive participation calculation: By taking a weighted average of the participation of sound data and the participation of image recognition, the impact of sound and image data on teaching quality is comprehensively considered to reduce the error of a single data source.

[0111] Based on the real-time collected data and evaluation results, the evaluation criteria and algorithms are dynamically adjusted to ensure that the system can adapt to different teaching environments and conditions and further improve accuracy.

[0112] Features are extracted from the sound decibel value and the student status classification results, and normalized so that each feature can be comprehensively evaluated on the same scale, reducing the impact of differences between features on the evaluation results.

[0113] Regularly adjust the evaluation criteria and feature weights to ensure that the comprehensive evaluation model can continue to reflect the actual teaching quality.

[0114] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. The method of evaluating the quality of ideological classroom teaching is characterized by: The following steps are involved: Step 1: Sound decibel detection: Use a microphone array to place multiple microphones at different locations in the classroom to separate and capture sounds from different sources, and use sound source localization and acoustic model analysis to reduce noise interference; Step 2: Student status detection: In a multi-camera system, deep learning algorithms are used for image recognition to accurately determine the student status: listening to a class, discussing, or sleeping. The trained deep learning model is used to analyze the student's posture and facial expression, accurately classify the student's status, and combine real-time lighting data and camera angle calibration to further improve the accuracy of status recognition. Step 3: Sound data processing: integrate the sound data collected by multiple microphones, apply filters and noise suppression technology to reduce the impact of environmental noise on the sound decibel value, and classify the teaching sound and discussion sound: calculate the decibel values ​​of the teaching sound and discussion sound respectively, and compare them with the reference values ​​in the storage module to obtain the sound-related teaching quality indicators; Step 4: Student status analysis: By processing students’ posture images in real time, applying image enhancement technology and image recognition algorithms, we can determine whether students are listening to a class, discussing, or sleeping. We can compare students’ status data with reference values, and combine image recognition results and sound data to evaluate students’ participation. Step 5: Comprehensive evaluation model: Combine the sound decibel value and student status information to calculate the teaching quality using a comprehensive evaluation model; Step 6: Dynamic adjustment: Adjust the evaluation criteria and algorithms based on the real-time collected data and evaluation results to ensure the accuracy and adaptability of the system; The original image in step 4 is I original , the image enhancement function is f, and the image after enhancement is I enhanced , the image enhancement expression is: I enhanced =f(I original ,transformations) Where: f(·) is the image enhancement function, which is used to process the original image and apply the specified transformation to generate the enhanced image I enhanced ; Suppose the student status prediction after CNN processing is in The category with the highest probability is: f CNN (I enhanced ) is the enhanced image I through the convolutional neural network enhanced Process and output a vector representing the probability of each category; The expression for comparing the state data with the reference value in step 4 is: Where: N is the total number of samples; The true label of the i-th sample; Accuracy is the accuracy, which calculates the proportion of all sample predictions that are correct, that is, the proportion of correctly predicted samples to the total number of samples; in is the Kronecker delta function:

2. The method for evaluating the quality of ideological classroom teaching according to claim 1 is characterized by: The expression combining the image recognition result and the sound data in step 4 is: The participation rate of sound data is P sound , the participation of image recognition is P image , comprehensive participation P total Calculated by weighted average: P total =α·P sound +(1-α)·P image Where: P sound is the participation of the sound data, indicating the participation degree or weight of the sound analysis results; α is the weight coefficient of the sound data, ranging from 0 to 1, which determines the influence ratio of the sound data in the calculation of the comprehensive participation; P total To synthesize engagement, the impact of sound data and image data was combined; Set the threshold as threshold, according to the comprehensive participation P total And threshold for final evaluation: Where: P total : Comprehensive participation, a comprehensive evaluation result combining sound and image data; S is the final evaluation result, based on P total Whether it is greater than or equal to the threshold determines the level of participation. If P total If it is greater than or equal to the threshold, it is evaluated as "high engagement"; otherwise, it is evaluated as "low engagement".

3. The method for evaluating the quality of ideological classroom teaching according to claim 2 is characterized by: The sound decibel detection expression in step 1 is: Where: L is the decibel value of the sound, indicating the sound pressure level; I0 is the reference sound intensity, which is taken as (1×10 -12 )W / m 2 , which is the minimum sound intensity that can be heard by the human ear.

4. The method for evaluating the quality of ideological classroom teaching according to claim 3 is characterized by: The sound source localization and acoustic model analysis expressions in step 1 are: Use a microphone array to locate multiple sound sources and calculate the direction θ and position (x, y, z) of the sound source, expressed as: in: Indicates the location of the sound source in three-dimensional space; The acoustic model M is used for analysis and calculation of the sound propagation characteristics, including reflection and refraction phenomena, which can be expressed mathematically as: Where t is the time variable; Reduce noise disturbance: The noise interference N is reduced by signal processing techniques and is defined as: S clean =S raw -N Where: S clean To remove the noise from the sound signal, S raw is the original sound signal.

5. The method for evaluating the quality of ideological classroom teaching according to claim 4 is characterized by: In step 2, the deep learning model is used for image classification to determine the status of the students. The expression is: state classification = argmax(f(I;⑩)) Where: I is the input image or video frame; ⑩ is the parameter of the deep learning model, including the weight of the convolution layer and the weight of the fully connected layer; f is the output function of the deep learning model, which returns the probability distribution of each state; argmax means selecting the state with the highest probability as the classification result; Suppose the output of the model is a vector y containing the probability of each state: y=[p 听课 ,p 讨论 ,p 睡觉 ] Then: state classification = argmax(p 听课 , p 讨论， p 睡觉 ).

6. The method for evaluating the quality of ideological classroom teaching according to claim 5 is characterized by: The calibration model expression in step 2 is: I 校准 =I·K(C) in: I 校准 is the calibrated image; I is the original image; K(C) is the camera calibration matrix function, which represents the transformation based on the camera configuration information: angle, focal length and position.

7. The method for evaluating the quality of ideological classroom teaching according to claim 6 is characterized by: In step 3, the sound data fusion: the sound data collected from multiple microphones needs to be fused to obtain a more accurate sound signal. i (t) represents the sound signal collected by the i-th microphone at time t, and the fused sound signal x(t) is expressed as: Where: N is the number of microphones; x i (t) is the signal collected by the i-th microphone at time t.

8. The method for evaluating the quality of ideological classroom teaching according to claim 7 is characterized by: The filter and noise suppression expressions in step 3 are: Assume that a linear filter is applied, and its filter coefficient is h(t). The signal after filtering is It is expressed as: Where: * represents the convolution operation, and the filter coefficient h(t) is designed as a low-pass filter or a band-pass filter; Noise suppression techniques include noise estimation and noise suppression filters, using spectral subtraction to reduce noise: in: is the estimated noise signal; Calculate the decibel value of the processed signal. Suppose the processed signal is The calculation formula of its decibel value l(t) is: Where: T is the time window length for calculating decibel values; It is the square value of the signal and represents the power of the signal.

9. The method for evaluating the quality of ideological classroom teaching according to claim 8 is characterized by: The feature extraction expression in step 5 is: The sound decibel value obtained from step 3 is L(t), and the student status classification result obtained from step 2 is S. The features include: sound decibel value: L(t); student status: S: S = listening to the class, S = discussing, or S = sleeping; Feature normalization: Standardize different features so that they can be integrated on the same scale. Let the maximum and minimum values ​​of L(t) be L max and l min , then the normalized decibel value l(t) is expressed as: The student status is converted into numerical features, including: listening to the class: 1; discussing: 2; sleeping: 0; Then the status information S norm Take these values ​​directly; The normalized features are input into the comprehensive evaluation model, and the weighted sum model is used to calculate the teaching quality index. The comprehensive evaluation Q is expressed as: Q=w1·l norm (t)+w2·S norm Among them: w1 and w2 are the weights of the features, indicating the contribution of each feature to the teaching quality; the setting of weights w1 and w2 is based on historical data or expert opinions, and the evaluation results are optimized through training models or adjustments.

10. The method for evaluating the quality of ideological classroom teaching according to claim 9 is characterized by: In step 6, the evaluation criteria and algorithm expressions are adjusted as follows: Adjust threshold: Q threshold ; Adjust weights: adjust feature weights w1 and w2 according to the importance of the features, and use regression analysis, optimization algorithms, or expert advice to determine new weights; assuming that the new weights are w′1 and w′2, the updated comprehensive evaluation model is: Q t =w′1·l norm (t)+w′2·S norm 。

Images