A teaching analysis method and system based on big data

Abstract

The present invention discloses a teaching analysis method and system based on big data, which relates to the field of teaching analysis technology, including: extracting features from the network scoring preprocessing data to obtain feature data; performing sample division on the feature data to obtain classification sample data; using the classification sample data to train and test a preset classification model, and obtaining a teaching level classification model through a deep learning method and in combination with a neural network model; obtaining the teaching level of the corresponding teacher based on the network scoring data of all students on the teacher and the teaching level classification model. The present invention can make full use of the network scoring data by combining the students' network scoring data on the teacher and the teaching level classification model, and can objectively evaluate the teacher's teaching level, providing an important reference basis for school management departments and individual teachers, and helping to improve teaching quality and effectiveness.

Description

Technical Field

[0001] The present invention relates to the technical field of teaching analysis, and in particular to a teaching analysis method and system based on big data. Background Art

[0002] At present, the teaching affairs systems of various universities have the function of student evaluation. The evaluation data stored in the student evaluation database amount to hundreds of thousands or even tens of millions. Moreover, as time goes by, various teaching activities begin to be carried out continuously, and these data are still increasing rapidly. The huge amount of data can be imagined.

[0003] However, such a huge amount of data is idle in the teaching service system of many universities and basically does not play any role. Even if it is used, it is basically a necessary condition for awards, evaluations, and professional title promotions. The data application is simplified, but this is only the tip of the iceberg of the true meaning and value of these data storages. It is far from playing its due value. The evaluation of teachers' teaching level only relies on students' scores for teachers and students' test scores. This not only fails to make full use of the teaching evaluation data in the teaching evaluation database, but also this evaluation method is too single and the evaluation results are not accurate enough, which is not conducive to improving teaching quality and effectiveness. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a teaching analysis method and system based on big data, aiming to solve the above-mentioned technical problems.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] In the first aspect, a teaching analysis method based on big data comprises the following steps:

[0007] Get the online rating data of all students for all teachers;

[0008] Preprocessing the network scoring data to obtain network scoring preprocessed data;

[0009] Performing feature extraction on the network scoring preprocessing data to obtain feature data;

[0010] Performing sample division on the feature data to obtain classified sample data;

[0011] The classification sample data is used to train and test the preset classification model, and a teaching level classification model is obtained through a deep learning method and combined with a neural network model;

[0012] Based on the online rating data of all students on teachers and the teaching level classification model, the teaching level of the corresponding teacher can be obtained.

[0013] Preferably, the network rating data includes a teaching attitude rating data set, a teacher quality rating data set, a teaching content rating data set, a teaching method rating data set and a teaching effect rating data set.

[0014] Preferably, preprocessing the network rating data to obtain network rating preprocessed data includes:

[0015] Identify samples with missing values ​​in network rating data;

[0016] According to each sample with missing values, the distance between the sample with missing values ​​and all complete samples is calculated by Euclidean distance;

[0017] According to the distance, determine the K samples closest to the sample as neighbors;

[0018] Based on the known values ​​of K neighbors, predict the missing value of the sample by taking the average;

[0019] Fill the predicted missing values ​​into the missing positions in the original data to obtain the filled data;

[0020] Convert the padded data to a uniform scale to obtain normalized data;

[0021] Features useful for analysis tasks are extracted from the normalized data to obtain network scoring preprocessing data.

[0022] Preferably, feature extraction is performed on the network scoring preprocessing data to obtain feature data, including:

[0023] Group the preprocessed network rating data according to user ID;

[0024] For each user's rating subset, calculate the rating data within a specific time period;

[0025] Based on the rating data, calculate the average rating of each user in a specific time period;

[0026] Based on the rating data, calculate the standard deviation of each user's rating in a specific time period;

[0027] The rating data of each user is sorted by timestamp, and the differences between adjacent ratings are calculated to obtain the rating trend of each user.

[0028] Preferably, according to each sample containing missing values, the distance between the sample with missing values ​​and all complete samples is calculated by Euclidean distance, including:

[0029] Determine a sample as a target sample from a sample set containing missing values;

[0030] For each complete sample in the dataset, calculate the distance between the sample and the target sample ,in, is the importance of feature i; is the scale factor of feature i, is a parameter between 0 and 1. is a tuning parameter, is the nonlinear index of feature i, and are the values ​​of sample x and sample y on the i-th feature respectively.

[0031] Preferably, according to the distance, determining K samples closest to the sample as neighbors includes:

[0032] Sort the distances in ascending order to obtain a distance list;

[0033] From the distance list, select the top K samples as the nearest neighbors of a given sample.

[0034] Preferably, the missing value of the sample is predicted by taking the average value based on the known values ​​of the K neighbors, including:

[0035] Prediction Sample Missing values ​​on feature f, sample The K nearest neighbors of , and the known values ​​on the feature f are , each neighbor To the sample The distance is , the corresponding weight is , then the sample Missing values ​​on feature f The calculation formula is:

[0036] ;

[0037] ;

[0038] in, is the width parameter of the Gaussian kernel.

[0039] Second, a teaching analysis system based on big data, including:

[0040] An acquisition module is used to acquire network rating data of all students on all teachers; preprocess the network rating data to obtain network rating preprocessed data; and perform feature extraction on the network rating preprocessed data to obtain feature data;

[0041] The processing module is used to perform sample division on the feature data to obtain classified sample data; use the classified sample data to train and test a preset classification model, and obtain a teaching level classification model through a deep learning method combined with a neural network model; and obtain the teaching level of the corresponding teacher based on the network rating data of all students on the teacher and the teaching level classification model.

[0042] According to a third aspect, a computing device includes:

[0043] one or more processors;

[0044] A storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement any of the methods described above.

[0045] In a fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, any of the methods described above is implemented.

[0046] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0047] By preprocessing the online rating data, including missing value processing and data filling, the completeness and accuracy of students' online rating data of teachers can be improved, thereby ensuring the reliability of subsequent analysis.

[0048] By grouping the data according to user ID and calculating features such as the average score, score standard deviation, and score trend within a specific time period, it helps to gain a deeper understanding of each student's evaluation habits and changing trends, thereby improving the accuracy of the teaching level classification model in classifying teachers' teaching levels to a certain extent.

[0049] By combining students' online rating data of teachers with the teaching level classification model, the online rating data can be fully utilized. At the same time, the teaching level of teachers can be objectively evaluated, providing important reference for school management departments and individual teachers, which will help improve the quality and effectiveness of teaching.

[0050] By analyzing the online rating data of all students on teachers, we can obtain the degree of influence of various factors on the teaching level of teachers, so that teachers in the whole school can adjust their teaching methods according to the degree of influence of various factors on the teaching level, thereby facilitating the optimization and improvement of teachers' teaching level, thereby achieving the goal of improving teaching quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0052] Figure 1 A flow chart of a teaching analysis method based on big data is shown.

[0053] Figure 2 A schematic diagram of a teaching analysis system based on big data is shown. DETAILED DESCRIPTION

[0054] 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.

[0055] like Figure 1 As shown, a teaching analysis method based on big data includes the following steps:

[0056] Step 11, obtaining the online rating data of all students for all teachers;

[0057] Step 12, preprocessing the network scoring data to obtain network scoring preprocessing data;

[0058] Step 13, extracting features from the network score preprocessing data to obtain feature data;

[0059] Step 14, performing sample division on the feature data to obtain classified sample data;

[0060] Step 15, using the classification sample data to train and test the preset classification model, through the deep learning method, and in combination with the neural network model, to obtain the teaching level classification model;

[0061] Step 16, according to the online rating data of all students on the teacher and the teaching level classification model, to obtain the teaching level of the corresponding teacher.

[0062] In an embodiment of the present invention, by preprocessing the network rating data, including missing value processing and data filling, the integrity and accuracy of the network rating data of students on teachers can be improved, thereby ensuring the reliability of subsequent analysis. By grouping the data according to user ID and calculating the average rating, rating standard deviation and rating trend in a specific time period, it is helpful to deeply understand the evaluation habits and changing trends of each student, thereby improving the accuracy of the teaching level classification model for the classification of teachers' teaching levels to a certain extent. By combining the network rating data of students on teachers and the teaching level classification model, the network rating data can be fully utilized, and the teaching level of teachers can be objectively evaluated, providing important reference basis for school management departments and individual teachers, which is helpful to improve the quality and effect of teaching. By analyzing the network rating data of all students on teachers, the degree of influence of various factors on the teaching level of teachers is obtained, so that teachers in the whole school can adjust their own teaching methods according to the degree of influence of various factors on the teaching level, and then conveniently optimize and improve the teaching level of teachers, so as to achieve the purpose of improving teaching quality.

[0063] In a preferred embodiment of the present invention, the network scoring data includes a teaching attitude scoring data set, a teacher quality scoring data set, a teaching content scoring data set, a teaching method scoring data set and a teaching effect scoring data set.

[0064] In an embodiment of the present invention, by collecting scoring data of multiple dimensions, including teaching attitude, teacher quality, teaching content, teaching methods and teaching effects, the method can comprehensively and meticulously evaluate the teaching quality of teachers. This multi-dimensional evaluation method helps to more accurately reflect the actual teaching level of teachers and avoid the one-sidedness of a single indicator. By using big data analysis methods, the influence of subjective bias on the evaluation results can be reduced. Because the data comes from the scores of a large number of students, the evaluation results are more objective and fair, and can better reflect the true teaching level of teachers. By analyzing the scoring data of different dimensions, educational institutions can identify in which aspects teachers perform well and which aspects need improvement, which helps teachers to improve their teaching ability in a targeted manner and thus improve the quality of teaching. By evaluating the teaching level of teachers, educational institutions can allocate teaching resources more reasonably, such as arranging high-level teachers to undertake more teaching tasks, providing targeted teacher training and development opportunities, etc., which helps to improve the overall teaching quality and student learning effects. Through online scoring, students can express their feedback and suggestions on teaching more directly, and teachers can also understand students' learning needs and opinions more promptly, which helps to establish a better teacher-student relationship and improve teaching satisfaction and learning experience. By analyzing large amounts of scoring data, teachers can discover new teaching trends and student needs, thereby inspiring teaching innovations, helping to promote improvements in teaching methods and updates in teaching content to adapt to the ever-changing educational environment and student needs.

[0065] In a preferred embodiment of the present invention, the above step 12 may include:

[0066] Step 121, identifying samples containing missing values ​​in the network scoring data;

[0067] Step 122, according to each sample containing missing values, calculate the distance between the sample containing the missing value and all complete samples by using the Euclidean distance;

[0068] Step 123, according to the distance, determine the K samples closest to the sample as neighbors;

[0069] Step 124, predicting the missing value of the sample by taking the average value based on the known values ​​of the K neighbors;

[0070] Step 125, filling the predicted missing values ​​into the missing positions in the original data to obtain filled data;

[0071] Step 126, converting the filled data to a uniform scale to obtain normalized data;

[0072] Step 127, extracting features useful for the analysis task from the normalized data to obtain network scoring preprocessing data.

[0073] In an embodiment of the present invention, through the missing value processing of steps 121 to 125, the missing values ​​in the network scoring data can be identified and processed, and the K nearest neighbor method based on Euclidean distance is used to predict and fill the missing values, which helps to maintain the integrity of the data and reduce the information loss and analysis bias caused by missing data. The normalization processing of step 126 converts the data to a unified scale, eliminates the influence of different dimensions or value ranges between different features, and makes the different features comparable. The feature extraction of step 127 selects features useful for the analysis task from the normalized data, which helps to reduce the interference of redundant information and noise on the analysis results. The extracted effective features can more accurately reflect the essential characteristics of the data and improve the accuracy and generalization ability of the teaching level classification model. The quality of the preprocessed data is higher, which can improve the training efficiency and performance of the subsequent classification model. The noise and redundant information removed during the preprocessing process reduce the difficulty of model learning, and the filled missing values ​​and normalization processing ensure the stability and consistency of the model input data. Through this series of data preprocessing steps, educational institutions can rely more on high-quality data to make decisions about teachers' teaching levels. This data-driven decision-making method helps improve the objectivity and accuracy of decisions.

[0074] In a preferred embodiment of the present invention, the above step 13 may include:

[0075] Step 131, grouping the pre-processed network rating data according to user ID;

[0076] Step 132, for each user's rating subset, calculating the rating data within a specific time period;

[0077] Step 133, calculating the average score of each user in a specific time period based on the score data, may include extracting all scores in the time period from the score subset of the user; using the extracted score data, calculating the average score of the user in the specific time period, where the average score is the sum of all scores divided by the number of scores;

[0078] Step 134, calculating the standard deviation of each user's score in a specific time period based on the score data;

[0079] Step 135 , sorting the rating data of each user by timestamp, and calculating the difference between adjacent ratings to obtain the rating trend of each user.

[0080] In an embodiment of the present invention, by grouping data according to user ID, and calculating the average score, score standard deviation and score trend in a specific time period, it is helpful to deeply understand the evaluation habits and changing trends of each student, thereby improving the accuracy of the teaching level classification model for the classification of the teacher's teaching level to a certain extent, and by calculating the average score and standard deviation of each user in a specific time period, the overall score level of the user and the fluctuation degree of the score can be understood. In addition, the score data is sorted by timestamp, and the difference between adjacent scores is calculated, which can reveal the user's score change trend and further analyze the user's behavior pattern and preference changes. These analysis results can help understand the user's scoring behavior and provide a useful reference for improving the teaching level and curriculum design. By monitoring the user's scoring trend, the teaching strategy and content can be adjusted in time to better meet the needs of the user and improve the teaching quality and user satisfaction.

[0081] In another preferred embodiment of the present invention, the above step 132 may include:

[0082] Step 1221 , select a specific user (assuming the user ID is userid) from the pre-processed network rating data.

[0083] Step 1222, determining a specific time period, for example, from time Tstart to time Tend;

[0084] Step 1223, extracting all the scores in the time period Tstart to Tend from the score subset of the user;

[0085] Step 1224, pass Calculate the weighted standard deviation of all scores in the time period Tstart to Tend ,in, is the i-th rating value in the data set; is the weight associated with the ith rating value, n is the total number of ratings in the dataset, is the jth rating value in the dataset, is the weight associated with the jth score value, i and j are both index variables in the summation symbol;

[0086] Step 1225, analyze the weighted standard deviation To assess the volatility of the score, if the weighted standard deviation is small, indicating that the user's ratings during this period are relatively consistent; if the weighted standard deviation If the value is large, it means that the user's rating fluctuates greatly during this period, and further analysis is needed.

[0087] In the embodiment of the present invention, step 1221 helps to understand the user's rating pattern and preferences in a personalized way. By analyzing a specific user, its rating habits, consistency, and possible biases or tendencies can be explored more deeply. Step 1222, determining a specific time period makes the analysis more targeted and timely. This helps to capture the changes in users' ratings in a specific period, such as seasonal trends, reactions before and after product updates, etc. In addition, it can also help eliminate interference from other irrelevant time periods, making the analysis more accurate. Step 1223, effectively filters the data set, retaining only the ratings related to the target user and time period, which greatly reduces the complexity and calculation of data analysis, while ensuring the focus and accuracy of the analysis. In this way, a clearer insight into the user's rating behavior within a specified time range can be obtained. Step 1224, calculating the weighted standard deviation takes into account the importance or reliability differences of different ratings, thereby providing a more accurate volatility assessment. This method is more reflective of the actual situation than the simple standard deviation because it can adjust its impact according to the weight of the rating, which helps to more accurately identify outliers or unstable factors in the rating data. Step 1225, by analyzing the weighted standard deviation, we can gain a deeper understanding of the stability and consistency of user ratings. A smaller weighted standard deviation indicates that users' ratings are relatively consistent over this period of time, which may mean that their views on the product or service are relatively stable, while a larger weighted standard deviation indicates that the ratings are more volatile, and we may need to further explore the reasons behind them, such as changes in user sentiment, fluctuations in product quality, etc. This analysis helps to better understand user behavior and market dynamics.

[0088] In a preferred embodiment of the present invention, the above step 122 may include:

[0089] Step 1221, determining a sample from the sample set containing missing values ​​as a target sample;

[0090] Step 1222, for each complete sample in the data set, calculate the distance between the sample and the target sample ,in, is the importance of feature i; is the scale factor of feature i, is a parameter between 0 and 1. is a tuning parameter, is the nonlinear index of feature i, and are the values ​​of sample x and sample y on the i-th feature respectively.

[0091] In the embodiment of the present invention, by determining a target sample from a sample set containing missing values ​​and calculating the distance between the complete sample and the target sample, the present invention does not need to directly fill in the missing values ​​or delete the samples containing missing values, thereby avoiding the possible introduction of noise or information loss. It represents the importance of feature i. By introducing feature importance weighting, the distance calculation formula can more accurately reflect the importance differences of different features in distance calculation. As the scale factor of feature i, it can adjust the scale of different features in distance calculation, which helps to deal with features with different dimensions or scales, making the distance calculation more reasonable and accurate. , and are adjustable parameters that provide flexibility to adjust the shape and sensitivity of the distance calculation. At the same time, these parameters are also interpretable and can help understand the impact of different features on the distance calculation. By introducing exponential functions and weight factors, the formula can suppress the impact of outliers or noise on the distance calculation to a certain extent, thereby improving the robustness of the calculation.

[0092] In a preferred embodiment of the present invention, the above step 123 may include:

[0093] Step 1231, sort the distances in ascending order to obtain a distance list;

[0094] Step 1232, select the first K samples from the distance list as the nearest neighbors of the given sample.

[0095] In an embodiment of the present invention, by sorting the distances in ascending order through step 1231, it can be ensured that the selected neighbors are the closest to the given sample in terms of distance, and the sorting process enables clear identification of which samples are most similar to the samples containing missing values. By selecting only the first K nearest neighbors (step 1232), the number of samples to be considered in subsequent steps can be reduced, which helps to reduce computational complexity and memory consumption, especially when processing large-scale data sets. When predicting missing values ​​using the K nearest neighbor method, selecting the nearest K neighbors can reduce the prediction bias caused by a single abnormal neighbor, and by comprehensively considering the information of multiple similar samples, a more robust and reliable missing value prediction result can be obtained. Since the selected neighbors are very close to the samples containing missing values ​​in the feature space, they are likely to have similar feature values ​​and scoring patterns, which makes it more reasonable and accurate to use the average values ​​of these neighbors to fill the missing values. By adjusting the value of K, the number of neighbors used to predict missing values ​​can be controlled. This provides flexibility for data analysis, allowing the selection of a suitable K value according to the characteristics of the data and the analysis requirements. For example, when the data is noisy, it may be more appropriate to select a smaller K value; while when the data distribution is sparse, it may be more helpful to select a larger K value.

[0096] In a preferred embodiment of the present invention, the above step 124 may include:

[0097] Step 1241, predicting samples Missing values ​​on feature f, sample The K nearest neighbors of , and the known values ​​on the feature f are , each neighbor To the sample The distance is , the corresponding weight is , then the sample Missing values ​​on feature f The calculation formula is:

[0098] ;

[0099] ;

[0100] in, is the width parameter of the Gaussian kernel.

[0101] In an embodiment of the present invention, by considering the distance between each neighbor and the sample and assigning a corresponding weight to each neighbor, the method can more accurately predict missing values. Closer neighbors will receive higher weights, thereby having a greater impact on the prediction results, which is in line with intuition and reality, because similar samples are more likely to have similar eigenvalues. By introducing a Gaussian kernel function to calculate weights, the method provides additional flexibility. The width parameter σ of the Gaussian kernel can be adjusted according to the characteristics of the data to optimize the prediction results. This flexibility enables the method to adapt to data sets of different densities and distributions. Compared with the traditional K nearest neighbor method, the weighted method has better robustness for outliers or noise data. Since outliers are usually far away from other samples, they will receive lower weights, thereby having less impact on the final prediction results. By comprehensively considering the information of multiple neighbors, the method can make more full use of available data, and by improving the accuracy of missing value prediction, the method helps to improve the quality and reliability of the entire data analysis process.

[0102] In another preferred embodiment of the present invention, the above step 14 may include:

[0103] Step 141, randomly divide the data set into a training set and a test set;

[0104] Step 142, determining the ratio of the training set to the test set;

[0105] Step 143, dividing the feature data according to the ratio of the training set to the test set;

[0106] Step 144, after the division is completed, training sample data and test sample data are obtained.

[0107] In an embodiment of the present invention, step 141, in this step, the entire data set is randomly shuffled to ensure that each data point has an equal chance of being assigned to the training set or the test set, and the randomization is to avoid any potential bias that may exist due to the original order of the data. Specifically, it includes using a random number generator or a corresponding software library function to shuffle the order of the data set to ensure that the randomization process is repeatable, which can be achieved by setting a random seed. Step 142, determine the size ratio of the training set and the test set, which is set according to the specific needs of the project and the total amount of available data. Specifically, it includes selecting a suitable ratio according to empirical rules or project requirements, in which the larger part is used for the training set and the smaller part is used for the test set; ensure that the selected ratio can provide enough training data to train the model, and also retain enough test data to accurately evaluate the performance of the model. Step 143, according to the previously determined ratio, the randomized data set is divided into a training set and a test set. Specifically, according to the ratio determined in step 142, the first part of the data set is allocated to the training set, and the remaining part of the data set is allocated to the test set, ensuring that the allocation process maintains the integrity and representativeness of the data, that is, each set should contain various types of data points in the original data set. Step 144, after completing the data partitioning, you will have two independent data sets: one for training the model (training set) and the other for evaluating the performance of the model (test set). Specifically, confirm that the partitioning of the training set and the test set is accurate, use the training set in the model training phase, and keep the test set so that it can be used to evaluate the model's accuracy, generalization ability and other indicators after the model training is completed; ensure that the training set and the test set remain separate during the model training and evaluation process to avoid data leakage and overfitting problems.

[0108] In another preferred embodiment of the present invention, after the above step 16, the following steps may also be included:

[0109] All students' online rating data on teachers are input into the teaching level classification model to obtain the teaching level score of each teacher; the teaching level score of each teacher is statistically analyzed to calculate the mean and standard deviation of each teacher's teaching level score.

[0110] Based on the calculated mean and standard deviation, the distribution of each teacher's teaching level score is determined, and outliers or outliers are identified; multiple teaching level levels are set, and corresponding score ranges are set for each level. The number of levels and score ranges can be determined based on actual needs and teaching evaluation standards.

[0111] Match each teacher's teaching level score with the set teaching level level to determine each teacher's teaching level; for teachers at a lower teaching level, further analyze the reasons and provide corresponding improvement suggestions or training opportunities to improve their teaching level; regularly update and optimize the teaching level classification model as needed to ensure the accuracy and applicability of the model.

[0112] In the embodiment of the present invention, the teaching level of each teacher can be evaluated more comprehensively and targeted improvement suggestions can be provided. At the same time, by regularly updating and optimizing the teaching level classification model, it can adapt to changes in different teaching environments and evaluation standards.

[0113] In another preferred embodiment of the present invention, after determining the teaching level of each teacher, the following steps may also be included:

[0114] Get the IDs of all teachers whose teaching level is excellent;

[0115] Search and obtain the corresponding teacher's teaching attitude rating value, teacher quality rating value, teaching content rating value, teaching method rating value, teaching effect rating value and teaching level rating value through the teacher's ID;

[0116] Establish a multiple linear regression model and use Solve each influence coefficient to analyze the impact of teaching attitude, teacher quality, teaching content, teaching method and teaching effect score on the teaching level score. represents the dependent variable, represents the intercept term, which represents the expected value of the dependent variable when all independent variables are 0; ( , , , , ) represents the coefficient of the independent variable, corresponding to ( , , , , ), which measures the direct effect of each variable on the dependent variable; represents the interaction term ( ) coefficient, which measures and The dependent variable The impact of and When changing at the same time, how it will change; Nonlinear term The coefficient of and There may be nonlinear relationships between represents the error term;

[0117] The solved influence coefficients are arranged in order from large to small to indicate the influence of each factor on the teacher's teaching level.

[0118] In an embodiment of the present invention, the IDs of all teachers with excellent teaching levels are obtained. This step can accurately locate the group of teachers with high teaching levels, and provide basic data for the subsequent analysis of the common characteristics and success factors of these teachers. By focusing only on excellent teachers, it is possible to more effectively identify which factors are essential for improving the teaching level to an excellent level. Collecting comprehensive and detailed teacher evaluation data is helpful to establish an accurate analysis model. These data can reflect the performance of teachers in different dimensions, so as to more comprehensively understand which factors affect the teaching level of teachers. The multivariate linear regression model can quantitatively analyze the impact of multiple independent variables on the dependent variable. By solving the influence coefficient, the specific contribution of each factor (such as teaching attitude, teacher quality, etc.) to the teaching level score can be clarified, which helps educators to more accurately understand which factors are the key to improving the teaching level. By sorting the influence coefficient, the order of importance of each factor on the teaching level can be intuitively seen, which helps education managers and teachers to give priority to and improve those factors that have the greatest impact on improving the teaching level under limited resources, so as to achieve the purpose of optimizing the allocation of teaching resources and improving teaching effects.

[0119] In another preferred embodiment of the present invention, the solved influence coefficients are arranged in order from large to small to indicate the influence of each factor on the teacher's teaching level, including:

[0120] Extract the influence coefficient of each independent variable from the multivariate linear regression model;

[0121] The random forest model is constructed using the random forest algorithm. The prediction accuracy is improved by constructing multiple decision trees and combining the prediction results of the random forest model. When constructing each tree, the random forest algorithm randomly extracts a portion of features and samples from the data set for training to increase the diversity and robustness of the model.

[0122] Use the organized data set to train the random forest model. During the training process, the random forest model will learn how to predict the target variable based on the input features. At the same time, the model will calculate the importance of each feature in the decision tree construction process.

[0123] After training is complete, the importance score of each feature can be extracted from the random forest model;

[0124] For each independent variable, calculate the weighted average of its influence coefficient and importance score. The weighted average can be calculated by the following formula:

[0125] ;

[0126] in, and are the weights of the influence coefficient of the multivariate linear regression model and the importance score of the feature importance evaluation algorithm, and ( + = 1), the weight can be determined according to the actual application scenario and the degree of trust in the two methods. is the independent variable in the multiple linear regression model The influence coefficient of is to extract each independent variable in the random forest model The importance score of Indicates all influence coefficients The maximum absolute value of Represents all importance scores The maximum value of

[0127] According to the calculated weighted average, all independent variables are sorted from large to small. The sorted list will show the relative importance of each factor on the numerical value of the teaching level score.

[0128] In an embodiment of the present invention, the influence coefficient of each independent variable is extracted from the multivariate linear regression model, which can quantify the direct influence of each independent variable (factor) on the teacher's teaching level, provide the statistical contribution of each factor to the target variable, and help to preliminarily understand which factors have a significant impact on the teaching level. As an integrated learning method, random forest can improve the accuracy and stability of prediction by constructing multiple decision trees and combining their prediction results. At the same time, randomly extracting features and samples for training increases the diversity of the model and the robustness to noise data. Through training, the random forest model can learn the relationship between the input features and the target variable, which provides a basis for subsequent feature importance evaluation. In addition, feature selection during training helps to reduce the impact of irrelevant or redundant features on model performance. The feature importance score reflects the contribution of each feature in the model, which helps to identify the key features that have a greater impact on the teaching level evaluation. These scores provide an important basis for subsequent feature selection, model optimization and interpretability. By calculating the weighted average, the influence coefficient of the multivariate linear regression model and the feature importance score of the random forest model can be comprehensively considered to obtain a more comprehensive independent variable importance evaluation result. This method combines the advantages of two different algorithms and improves the accuracy and reliability of the evaluation. The sorted list clearly shows the relative importance of each factor in affecting the teaching level score, which helps decision makers quickly identify key factors, formulate targeted improvement strategies and optimize resource allocation. At the same time, this sorting result also provides valuable reference information for subsequent research and analysis.

[0129] A teaching analysis system 20 based on big data, comprising:

[0130] The acquisition module 21 is used to obtain the network rating data of all students on all teachers; pre-process the network rating data to obtain network rating pre-processed data; perform feature extraction on the network rating pre-processed data to obtain feature data;

[0131] The processing module 22 is used to perform sample division on the feature data to obtain classified sample data; use the classified sample data to train and test the preset classification model, and obtain the teaching level classification model through the deep learning method and in combination with the neural network model; obtain the teaching level of the corresponding teacher based on the network rating data of all students on the teacher and the teaching level classification model.

[0132] Optionally, the network rating data includes a teaching attitude rating data set, a teacher quality rating data set, a teaching content rating data set, a teaching method rating data set and a teaching effect rating data set.

[0133] Optionally, preprocessing the network scoring data to obtain network scoring preprocessed data includes:

[0134] Identify samples with missing values ​​in network rating data;

[0135] According to each sample with missing values, the distance between the sample with missing values ​​and all complete samples is calculated by Euclidean distance;

[0136] According to the distance, determine the K samples closest to the sample as neighbors;

[0137] Based on the known values ​​of K neighbors, predict the missing value of the sample by taking the average;

[0138] Fill the predicted missing values ​​into the missing positions in the original data to obtain the filled data;

[0139] Convert the padded data to a uniform scale to obtain normalized data;

[0140] Features useful for analysis tasks are extracted from the normalized data to obtain network scoring preprocessing data.

[0141] Optionally, feature extraction is performed on the network score preprocessing data to obtain feature data, including:

[0142] Group the preprocessed network rating data according to user ID;

[0143] For each user's rating subset, calculate the rating data within a specific time period;

[0144] Based on the rating data, calculate the average rating of each user in a specific time period;

[0145] Based on the rating data, calculate the standard deviation of each user's rating in a specific time period;

[0146] The rating data of each user is sorted by timestamp, and the differences between adjacent ratings are calculated to obtain the rating trend of each user.

[0147] Optionally, for each sample with missing values, the distance between the sample with missing values ​​and all complete samples is calculated using the Euclidean distance, including:

[0148] Determine a sample as a target sample from a sample set containing missing values;

[0149] For each complete sample in the dataset, calculate the distance between the sample and the target sample ,in, is the importance of feature i; is the scale factor of feature i, is a parameter between 0 and 1. is a tuning parameter, is the nonlinear index of feature i, and are the values ​​of sample x and sample y on the i-th feature respectively.

[0150] Optionally, based on the distance, determine the K samples closest to the sample as neighbors, including:

[0151] Sort the distances in ascending order to obtain a distance list;

[0152] From the distance list, select the top K samples as the nearest neighbors of a given sample.

[0153] Optionally, predict the missing value of the sample by taking the average based on the known values ​​of the K neighbors, including:

[0154] Prediction Sample Missing values ​​on feature f, sample The K nearest neighbors of , and the known values ​​on the feature f are , each neighbor To the sample The distance is , the corresponding weight is , then the sample Missing values ​​on feature f The calculation formula is:

[0155] ;

[0156] ;

[0157] in, is the width parameter of the Gaussian kernel.

[0158] It should be noted that the system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0159] An embodiment of the present invention further provides a computing device, comprising: a processor, a memory storing a computer program, wherein the computer program, when executed by the processor, executes the method described above. All implementations in the above method embodiments are applicable to this embodiment, and can achieve the same technical effect. An embodiment of the present invention further provides a computer-readable storage medium, storing instructions, which, when executed on a computer, enables the computer to execute the method described above. All implementations in the above method embodiments are applicable to this embodiment, and can achieve the same technical effect.

[0160] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0161] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0162] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0163] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0164] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0165] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, ROM, RAM, magnetic disks, or optical disks.

[0166] In addition, it should be pointed out that in the apparatus and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. Moreover, the steps of performing the above series of processing can naturally be performed in chronological order according to the order of description, but it is not necessary to perform them in chronological order, and some steps can be performed in parallel or independently of each other. For those of ordinary skill in the art, it can be understood that all or any steps or components of the method and apparatus of the present invention can be implemented in hardware, firmware, software or a combination thereof in any computing device (including processors, storage media, etc.) or a network of computing devices, which can be achieved by those of ordinary skill in the art using basic programming skills after reading the description of the present invention.

[0167] Therefore, the purpose of the present invention can also be achieved by running a program or a group of programs on any computing device. The computing device can be a well-known general device. Therefore, the purpose of the present invention can also be achieved by simply providing a program product containing a program code that implements the method or device. That is to say, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any well-known storage medium or any storage medium developed in the future. It should also be pointed out that in the device and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. In addition, the steps of performing the above-mentioned series of processing can naturally be performed in chronological order according to the order of description, but it is not necessary to perform them in chronological order. Some steps can be performed in parallel or independently of each other.

[0168] The above description of the embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A teaching analysis method based on big data, characterized in that: The following steps are involved: Get the online rating data of all students for all teachers; Preprocessing the network scoring data to obtain network scoring preprocessed data; Performing feature extraction on the network scoring preprocessing data to obtain feature data; Performing sample division on the feature data to obtain classified sample data; The classification sample data is used to train and test the preset classification model, and a teaching level classification model is obtained through a deep learning method and combined with a neural network model; Based on the online rating data of all students on teachers and the teaching level classification model, the teaching level of the corresponding teacher is obtained; The network scoring data includes a teaching attitude scoring data set, a teacher quality scoring data set, a teaching content scoring data set, a teaching method scoring data set and a teaching effect scoring data set; Preprocessing the network scoring data to obtain network scoring preprocessed data includes: Identify samples with missing values ​​in network rating data; According to each sample with missing values, the distance between the sample with missing values ​​and all complete samples is calculated by Euclidean distance; According to the distance, determine the K samples closest to the sample as neighbors; Based on the known values ​​of K neighbors, predict the missing value of the sample by taking the average; Fill the predicted missing values ​​into the missing positions in the original data to obtain the filled data; Convert the padded data to a uniform scale to obtain normalized data; Extract features useful for analysis tasks from normalized data to obtain network scoring preprocessing data; Performing feature extraction on the network scoring preprocessing data to obtain feature data includes: Group the preprocessed network rating data according to user ID; For each user's rating subset, calculate the rating data within a specific time period; Based on the rating data, calculate the average rating of each user in a specific time period; Based on the rating data, calculate the standard deviation of each user's rating in a specific time period; Sort each user's rating data by timestamp and calculate the difference between adjacent ratings to get the rating trend of each user; According to each sample with missing values, the distance between the sample with missing values ​​and all complete samples is calculated by Euclidean distance, including: Determine a sample as a target sample from a sample set containing missing values; For each complete sample in the dataset, calculate the distance between the sample and the target sample ,in, is the importance of feature i; is the scale factor of feature i, is a parameter between 0 and 1. is a tuning parameter, is the nonlinear index of feature i, and are the values ​​of sample x and sample y on the i-th feature respectively; According to the distance, the K samples closest to the sample are determined as neighbors, including: Sort the distances in ascending order to obtain a distance list; From the distance list, select the first K samples as the nearest neighbors of a given sample; Based on the known values ​​of the K neighbors, the missing value of the sample is predicted by taking the average, including: Prediction Sample Missing values ​​on feature f, sample The K nearest neighbors of , and the known values ​​on the feature f are , each neighbor To the sample The distance is , the corresponding weight is , then the sample Missing values ​​on feature f The calculation formula is: ； in, is the width parameter of the Gaussian kernel, ; For each user's rating subset, calculate the rating data within a specific time period, which can include: Select a specific user from the preprocessed network rating data, the user ID is userid; Determine a specific time period, from time Tstart to time Tend; Extract all the ratings in the time period Tstart to Tend from the rating subset of the user; pass Calculate the weighted standard deviation of all scores in the time period Tstart to Tend ,in, is the i-th rating value in the data set; is the weight associated with the ith rating value, n is the total number of ratings in the dataset, is the jth rating value in the dataset, is the weight associated with the jth score value, i and j are both index variables in the summation symbol; Analyze weighted standard deviation To assess the volatility of the ratings.

2. A teaching analysis system based on big data, characterized in that: The system is used to perform the method according to claim 1, and the system comprises: An acquisition module is used to acquire network rating data of all students on all teachers; preprocess the network rating data to obtain network rating preprocessed data; and perform feature extraction on the network rating preprocessed data to obtain feature data; The processing module is used to perform sample division on the feature data to obtain classified sample data; use the classified sample data to train and test a preset classification model, and obtain a teaching level classification model through a deep learning method combined with a neural network model; and obtain the teaching level of the corresponding teacher based on the network rating data of all students on the teacher and the teaching level classification model.

3. A computing device, characterized in that: include: one or more processors; A storage device, used for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in claim 1.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which implements the method according to claim 1 when executed by a processor.

Images