Ideological and political education effect evaluation method based on Internet of Things

Abstract

The invention provides an ideological and political education effect evaluation method based on the Internet of Things, and the method comprises the steps: collecting student behavior data through a multi-modal sensing device, generating a standardized behavior data set through preprocessing and fusion, constructing classroom participation, learning autonomy and social practice three-dimensional feature vectors, and inputting a pre-training evaluation model. And outputting a quantitative score and a qualitative label, and realizing intelligent evaluation and feedback of an ideological and political education effect in combination with visual display and personalized improvement of a suggestion report. Scientific basis can be provided for personalized teaching improvement, and the quality and pertinence of ideological and political education are improved.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology and ideological and political education evaluation, and more specifically, to an IoT-based method for evaluating the effectiveness of ideological and political education. Background Technology

[0002] In the field of ideological and political education, traditional assessment methods mainly rely on teachers' subjective evaluations, students' self-reports, and simple exam scores. These methods are insufficient to comprehensively and objectively reflect students' true performance and learning outcomes in ideological and political education. For example, teachers' subjective evaluations may be influenced by personal preferences, and exam scores only reflect students' mastery of knowledge, failing to reflect the multi-dimensional effects of ideological and political education, such as students' classroom participation, self-discipline, and social practice abilities. Furthermore, existing assessment methods often lack real-time and dynamic aspects, failing to provide timely feedback on students' learning status and thus failing to meet the needs of personalized teaching.

[0003] With the rapid development of IoT technology, various sensing devices can collect students' behavioral data in real time, providing new possibilities for evaluating the effectiveness of ideological and political education. However, there is currently no mature technical solution that can fully utilize IoT multimodal data to conduct a comprehensive, objective, and dynamic evaluation of the effectiveness of ideological and political education.

[0004] In implementing the embodiments of the present invention, the prior art has at least the following problems or defects: traditional evaluation methods are highly subjective and have a single dimension, which cannot fully reflect the multi-dimensional effects of ideological and political education; they lack real-time and dynamic characteristics, making it difficult to meet the timely feedback needs of personalized teaching; and they have not effectively integrated Internet of Things technology, failing to make full use of multimodal data for intelligent evaluation. Summary of the Invention

[0005] This invention provides a method for evaluating the effectiveness of ideological and political education based on the Internet of Things, including:

[0006] S1: Collect students' multimodal behavior data through multiple IoT sensing devices deployed in ideological and political education scenarios, and upload the multimodal behavior data to the central processing server;

[0007] S2: The central processing server preprocesses and fuses the received multimodal behavior data to generate a standardized student behavior dataset;

[0008] S3: Based on the standardized student behavior dataset, construct a multi-dimensional feature vector of ideological and political education effectiveness, wherein the feature vector of ideological and political education effectiveness includes classroom participation dimension features, learning self-discipline dimension features and social practice dimension features;

[0009] S4: Input the feature vector of the ideological and political education effect into a pre-trained ideological and political education effect evaluation model, and the ideological and political education effect evaluation model outputs a quantitative comprehensive evaluation score and a qualitative evaluation label.

[0010] S5: Display the comprehensive evaluation score and the qualitative evaluation label through a visual interface, and generate a personalized report of suggestions for improving ideological and political education.

[0011] Furthermore, step S1 specifically includes:

[0012] S101: Collect students' facial image data and body movement video data in the classroom through the camera in the IoT sensing device;

[0013] S102: Collect students' classroom voice interaction data through the microphone array in the IoT sensing device;

[0014] S103: Collect students' heart rate variability data and physical activity intensity data through the smart bracelet in the IoT sensing device;

[0015] S104: Collect student classroom attendance time data through the access card reader in the IoT sensing device;

[0016] S105: Collect log data of students accessing the ideological and political education digital resource platform through the students' personal smart terminal devices.

[0017] Furthermore, the data preprocessing and fusion described in step S2 specifically includes:

[0018] S201: Perform data cleaning on the multimodal behavior data to remove abnormal and noisy data;

[0019] S202: Perform timestamp alignment on the cleaned multimodal behavioral data to unify all data into the same time coordinate system;

[0020] S203: Perform data normalization processing on the timestamp-aligned multimodal behavior data to make all data have the same numerical dimension;

[0021] S204: The normalized multimodal behavior data is associated and integrated according to student identity identifiers to generate the standardized student behavior dataset for each student.

[0022] Furthermore, in step S3, the construction of a multi-dimensional feature vector of ideological and political education effectiveness specifically includes:

[0023] S301: Based on the standardized student behavior dataset, calculate the classroom participation dimension features, which include head-up rate, interaction response frequency, and facial focus index.

[0024] S302: Based on the standardized student behavior dataset, calculate the learning self-discipline dimension features, which include attendance rate, on-time homework submission rate, and extracurricular resource learning time.

[0025] S303: Based on the standardized student behavior dataset, calculate the social practice dimension features, which include the number of volunteer service check-ins and the percentage of effective speeches in group collaboration projects.

[0026] Furthermore, in step S301, the calculation methods for the head-up rate, the interaction response frequency, and the facial focus index are as follows:

[0027] The head-up rate ( From the formula Calculations show that

[0028] in, This indicates the number of student head-up image frames captured within a preset time period. This represents the total number of student image frames collected within the preset time period;

[0029] The interaction response frequency (R) is given by the formula Calculations show that

[0030] in, This indicates the number of valid verbal responses that were recognized from a student during class. Indicates the total class time;

[0031] The facial focus index (F) is calculated using the formula...

[0032]

[0033] The calculation shows that, Indicates the total number of sampling periods. This represents the heart rate variability data at time t. This represents the student's baseline heart rate variability data at rest.

[0034] Furthermore, in step S4, the pre-trained ideological and political education effectiveness evaluation model is obtained through the following steps:

[0035] S401: Collect feature vectors of the ideological and political education effects of multiple students within a historical time period as a training sample set;

[0036] S402: Obtain the real effect score and real effect level corresponding to the training sample set, as evaluated by ideological and political education experts, and use them as training labels;

[0037] S403: Using the training sample set as input and the training labels as expected output, supervise the training of the machine learning model until the model loss function converges to obtain the ideological and political education effect evaluation model.

[0038] Further, in step S403, the machine learning model is a gradient boosting decision tree model; the gradient boosting decision tree model minimizes a loss function by iteratively adding decision trees, and the model output of the k-th iteration... Represented as:

[0039]

[0040] in, This represents the predicted value of the i-th sample after the k-th iteration. This represents the predicted value of the i-th sample after the (k-1)-th iteration. The learning rate is used to control the contribution of each decision tree. This represents the decision tree model added in the k-th iteration. Let represent the feature vector of the ideological and political education effect of the i-th sample;

[0041] The loss function is the mean squared error function, with a regularization term added. Its formula is:

[0042]

[0043] in, This represents the value of the loss function. This indicates the number of samples in the training sample set. This represents the true performance score of the i-th sample. This represents the model's prediction score for the i-th sample. This represents the total number of leaf nodes in the model. This represents the weight value of the j-th leaf node. and This represents the regularization parameter that controls the complexity of the model.

[0044] Further, in step S4, the qualitative evaluation label is obtained by mapping the comprehensive evaluation score; the mapping rule is as follows: when the comprehensive evaluation score is in the interval [90, 100], the qualitative evaluation label is excellent; when the comprehensive evaluation score is in the interval [75, 90), the qualitative evaluation label is good; when the comprehensive evaluation score is in the interval [60, 75), the qualitative evaluation label is qualified; when the comprehensive evaluation score is in the interval [0, 60), the qualitative evaluation label is to be improved.

[0045] Furthermore, in step S5, generating a personalized report with suggestions for improving ideological and political education specifically includes:

[0046] S501: Compare the scores of each dimension of the ideological and political education effect feature vector with a preset threshold to identify the weak dimensions that are below the threshold.

[0047] S502: Based on a predefined knowledge base of improvement strategies, match at least one improvement suggestion for each of the aforementioned weakness dimensions;

[0048] S503: Summarize and structure all the matched improvement suggestions to generate the personalized ideological and political education improvement suggestion report.

[0049] Further, in step S4, the ideological and political education effectiveness evaluation model is a long short-term memory network model based on an attention mechanism; the process by which the long short-term memory network model based on an attention mechanism processes the temporally sequenced feature vector of the ideological and political education effectiveness includes:

[0050] S1001: Feature vector of ideological and political education effectiveness arranged in time series Input the data into a Long Short-Term Memory (LSTM) network layer to obtain the hidden state at each time step. ;

[0051] in, This represents the input feature vector sequence. This represents the feature vector at time step t. Represents the hidden state sequence, This represents the hidden state at time step t.

[0052] S1002: Calculate the attention weights of the hidden state at each time step through the attention layer. The calculation formula is as follows:

[0053]

[0054] in, This represents the attention weight at the i-th time step. It is an alignment model function. , , This is the trainable parameter matrix for the attention mechanism. This represents the hidden state at the i-th time step. This represents the hidden state at the last time step;

[0055] S1003: Perform a weighted summation on the hidden state sequence to obtain the context vector.

[0056] in, This represents a context vector that encapsulates important information from all time steps.

[0057] S1004: Pass the context vector c through a fully connected output layer to obtain the final comprehensive evaluation score.

[0058] The embodiments of the present invention have at least the following beneficial effects:

[0059] 1. By deploying various IoT sensors in ideological and political education settings, multimodal behavioral data of students is collected, enabling comprehensive and real-time monitoring of student behavior. This method not only captures students' classroom performance, such as facial expressions, body movements, and voice interactions, but also records physiological data such as heart rate variability and physical activity intensity, as well as extracurricular learning behavior and attendance. This provides a rich and multidimensional data foundation for evaluating the effectiveness of ideological and political education, solving the problems of single data sources and limited dimensions in traditional evaluation methods, making the evaluation results more comprehensive and objective.

[0060] 2. Data preprocessing and fusion techniques are employed to clean, timestamp-align, and normalize the collected multimodal behavioral data, which is then integrated according to student identification identifiers to generate a standardized student behavior dataset. This process effectively eliminates outlier and noisy data, unifies the data's time reference and numerical units, and improves data quality and usability. This provides high-quality data support for subsequent feature vector construction and evaluation model training, solving the problem of inaccurate evaluation results caused by non-standard data processing and poor data quality in existing technologies, and ensuring the stability and reliability of the evaluation model.

[0061] 3. A multi-dimensional feature vector of ideological and political education effectiveness is constructed, and a pre-trained assessment model is used to output quantitative scores and qualitative labels, while simultaneously generating personalized improvement suggestion reports. This method not only quantifies the effectiveness of ideological and political education but also visually displays the assessment results through a graphical interface, helping educators quickly understand students' learning status and problems. Furthermore, the personalized improvement suggestion reports provide teachers with targeted directions for teaching improvement, addressing the lack of feedback mechanisms and personalized guidance in traditional assessment methods, and improving the relevance and effectiveness of ideological and political education. Attached Figure Description

[0062] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. Several embodiments of the invention are illustrated in the drawings by way of example and not limitation, wherein:

[0063] Figure 1 This is a flowchart illustrating an Internet of Things-based method for evaluating the effectiveness of ideological and political education, as provided in an embodiment of the present invention. Detailed Implementation

[0064] The principles and spirit of the invention will now be described with reference to several exemplary embodiments. It should be understood that these embodiments are provided merely to enable those skilled in the art to better understand and implement the invention, and are not intended to limit the scope of the invention in any way. Rather, these embodiments are provided to make the invention more thorough and complete, and to fully convey the scope of the invention to those skilled in the art.

[0065] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0066] It should be noted that the number of any elements in the accompanying drawings is for illustrative purposes only and not as a limitation, and any naming is for distinction only and has no limiting meaning.

[0067] The following is for reference. Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the Internet of Things-based method for evaluating the effectiveness of ideological and political education. Figure 1 As shown, an evaluation method for the effectiveness of ideological and political education based on the Internet of Things includes:

[0068] S1: Collect students' multimodal behavior data through multiple IoT sensing devices deployed in ideological and political education scenarios, and upload the multimodal behavior data to the central processing server;

[0069] S2: The central processing server preprocesses and fuses the received multimodal behavior data to generate a standardized student behavior dataset;

[0070] S3: Based on the standardized student behavior dataset, construct a multi-dimensional feature vector of ideological and political education effectiveness, wherein the feature vector of ideological and political education effectiveness includes classroom participation dimension features, learning self-discipline dimension features and social practice dimension features;

[0071] S4: Input the feature vector of the ideological and political education effect into a pre-trained ideological and political education effect evaluation model, and the ideological and political education effect evaluation model outputs a quantitative comprehensive evaluation score and a qualitative evaluation label.

[0072] S5: Display the comprehensive evaluation score and the qualitative evaluation label through a visual interface, and generate a personalized report of suggestions for improving ideological and political education.

[0073] In this invention, multiple IoT sensing devices deployed in ideological and political education scenarios collect multimodal behavioral data of students and upload this data to a central processing server. The multimodal behavioral data referred to here includes various behavioral data of students during the ideological and political education process collected through different types of sensors, including images, audio, and physiological data. Cameras can capture students' facial expressions and body movements, microphones can record students' voice interactions, and smart bracelets can monitor students' heart rate changes. The central processing server is responsible for receiving and storing this data, providing a foundation for subsequent data processing and analysis. This data collection method comprehensively covers students' classroom performance and extracurricular behavior, providing a rich data source for evaluating the effectiveness of ideological and political education.

[0074] Specifically, IoT sensing devices include cameras, microphone arrays, smart bracelets, access card readers, and students' personal smart terminal devices. Cameras are used to collect facial image data and body movement video data of students in the classroom, which can reflect students' attention and participation. Microphone arrays are used to collect students' voice interaction data in the classroom, analyzing the frequency and content of students' speeches in classroom discussions. Smart bracelets collect students' heart rate variability data and physical activity intensity data, which can indirectly reflect students' emotional state and learning stress.

[0075] Access card readers are used to record students' class attendance time data to ensure that students attend classes on time. Personal smart terminal devices are used to collect log data of students' access to the ideological and political education digital resource platform, including study time, access frequency, etc., which can reflect students' learning self-discipline. The data collected by these devices can comprehensively cover multiple dimensions of students' classroom performance, learning self-discipline and social practice in ideological and political education, providing multi-dimensional data support for subsequent evaluation.

[0076] Preferably, in the data preprocessing and fusion stage, the collected multimodal behavioral data is first cleaned to remove abnormal and noisy data. For example, for image data captured by a camera, blurry or irrelevant image frames can be removed using image recognition algorithms; for audio data captured by a microphone, background noise can be removed. Next, the cleaned data undergoes timestamp alignment to unify all data into the same time coordinate system, ensuring temporal consistency across different modalities.

[0077] The timestamps of the camera image frames were aligned with the timestamps of the microphone audio to accurately match students' facial expressions and speech content in subsequent analysis. Then, the timestamp-aligned data was normalized to ensure all data points were on the same numerical scale, facilitating subsequent analysis and modeling. Finally, the normalized data was associated and integrated according to student identifiers to generate a standardized student behavior dataset for each student. This process not only improved data quality but also provided a high-quality data foundation for subsequent feature vector construction and model training evaluation.

[0078] In some embodiments, step S1 specifically includes:

[0079] S101: Collect students' facial image data and body movement video data in the classroom through the camera in the IoT sensing device;

[0080] S102: Collect students' classroom voice interaction data through the microphone array in the IoT sensing device;

[0081] S103: Collect students' heart rate variability data and physical activity intensity data through the smart bracelet in the IoT sensing device;

[0082] S104: Collect student classroom attendance time data through the access card reader in the IoT sensing device;

[0083] S105: Collect log data of students accessing the ideological and political education digital resource platform through the students' personal smart terminal devices.

[0084] The system collects diverse data on students' facial images and body movements in the classroom through cameras in IoT sensors, voice interaction data through microphone arrays, heart rate variability and physical activity intensity data through smart bracelets, attendance time data through access card readers, and log data of students' access to the ideological and political education digital resource platform through their personal smart devices. These devices capture a wide range of data, covering students' external behavior and internal physiological responses, reflecting their participation and learning status in the ideological and political education process from multiple dimensions. For example, facial image data can be used to analyze students' focus and emotional state, voice interaction data can reflect their classroom participation and interaction, and heart rate variability data can indirectly reflect their emotional stress and learning status.

[0085] Cameras in IoT sensing devices are used to collect facial images and body movement video data of students in the classroom. This data can be analyzed using image recognition technology to determine students' facial expressions and body movements; for example, by detecting whether students look up or smile, their level of focus and emotional state can be assessed. Microphone arrays are used to collect students' voice interaction data in the classroom. This data can be analyzed using speech recognition technology to determine the frequency and content of students' speech, thereby evaluating their classroom participation.

[0086] Smart bracelets collect students' heart rate variability and physical activity intensity data. Heart rate variability reflects students' emotional stress levels, while physical activity intensity reflects their classroom activity levels. Access card readers record students' class attendance time data, which can be used to calculate attendance rates and thus assess their learning self-discipline. Students' personal smart devices collect log data on their access to the ideological and political education digital resource platform. This data can include students' study time, access frequency, and study paths, thereby assessing students' self-discipline and self-learning ability outside of class.

[0087] Preferably, to ensure the accuracy and reliability of the data, the collected data can be further processed and analyzed. For example, for facial image data collected by a camera, deep learning algorithms can be used for facial recognition and expression analysis, and a trained model can be used to identify students' focus and emotional state. For voice interaction data collected by a microphone array, speech recognition technology can be used to convert speech into text, and then natural language processing technology can be used to analyze the content and frequency of students' speech. For heart rate variability data collected by a smart bracelet, noise can be removed using signal processing technology, and then statistical parameters of heart rate variability, such as standard deviation and mean, can be calculated to reflect students' emotional stress levels. For attendance time series data collected by an access card reader, students' attendance rate and the number of times they are late or leave early can be calculated to assess their learning self-discipline.

[0088] Log data collected from personal smart devices can be analyzed to examine students' study time, access frequency, and learning paths, thereby assessing their extracurricular learning behavior and self-learning abilities. These processing steps further improve data quality and analytical accuracy, providing more reliable data support for subsequent evaluations of ideological and political education effectiveness.

[0089] In some embodiments, the data preprocessing and fusion described in step S2 specifically includes:

[0090] S201: Perform data cleaning on the multimodal behavior data to remove abnormal and noisy data;

[0091] S202: Perform timestamp alignment on the cleaned multimodal behavioral data to unify all data into the same time coordinate system;

[0092] S203: Perform data normalization processing on the timestamp-aligned multimodal behavior data to make all data have the same numerical dimension;

[0093] S204: The normalized multimodal behavior data is associated and integrated according to student identity identifiers to generate the standardized student behavior dataset for each student.

[0094] In this invention, step S2 involves preprocessing and fusing the collected multimodal behavioral data to generate a standardized student behavior dataset. This process is crucial for ensuring data quality because the raw data often contains noise, outliers, and temporal inconsistencies between different devices. The goal of data preprocessing and fusing is to clean, align, and normalize this data so that feature vectors can be accurately constructed and evaluated subsequently.

[0095] Specifically, data cleaning removes outliers and noise from the data, timestamp alignment unifies data from different modalities onto the same time base, data normalization transforms the data into the same numerical range, and data association and integration integrates all the processed data according to student identity identifiers to form a dataset for each student.

[0096] Specifically, data cleaning refers to identifying and removing outliers and noisy data from the collected data using algorithms. For example, for image data, blurry or incomplete image frames can be removed by detecting image sharpness and integrity; for audio data, background noise can be removed using signal processing techniques. Secondly, timestamp alignment refers to unifying data from different modalities onto the same time base. For example, aligning the timestamps of image data captured by a camera with the timestamps of audio data captured by a microphone ensures temporal consistency between image and audio data. Next, data normalization refers to converting data from different modalities to the same numerical range; for example, normalizing heart rate variability data and physical activity intensity data to the range of 0 to 1 for subsequent analysis and processing. Finally, data association and integration refers to integrating all processed data according to student identity identifiers to form a standardized student behavior dataset for each student. This process ensures that the behavioral data of each student is logically complete and can be used for subsequent feature vector construction and model training evaluation.

[0097] Preferably, in the data cleaning stage, statistical methods can be used to identify outliers. For example, the mean and standard deviation of each data modality can be calculated, and data points exceeding a certain multiple of the standard deviation can be considered outliers and removed. In the timestamp alignment stage, interpolation methods can be used to fill in inconsistencies in timestamps. For example, for missing time points, linear interpolation or spline interpolation can be used to estimate their data values. In the data normalization stage, min-max normalization can be used to transform the values ​​of each data modality to the range of 0 to 1. In the data association and integration stage, a database management system can be used to store and manage the behavioral data of each student, ensuring the integrity and consistency of the data. These refined steps can further improve the quality and usability of the data, providing a more reliable data foundation for subsequent evaluation of the effectiveness of ideological and political education.

[0098] In some embodiments, step S3, constructing a multi-dimensional feature vector of ideological and political education effectiveness, specifically includes:

[0099] S301: Based on the standardized student behavior dataset, calculate the classroom participation dimension features, which include head-up rate, interaction response frequency, and facial focus index.

[0100] S302: Based on the standardized student behavior dataset, calculate the learning self-discipline dimension features, which include attendance rate, on-time homework submission rate, and extracurricular resource learning time.

[0101] S303: Based on the standardized student behavior dataset, calculate the social practice dimension features, which include the number of volunteer service check-ins and the percentage of effective speeches in group collaboration projects.

[0102] The feature vector includes features in classroom participation, self-discipline, and social practice, aiming to comprehensively evaluate the effectiveness of ideological and political education from multiple perspectives. These features can more accurately reflect students' performance and learning status during the ideological and political education process. For example, classroom participation can assess students' engagement in ideological and political education through classroom interaction; self-discipline can assess their self-learning ability through attendance and study habits; and social practice can assess their ability to apply ideological and political knowledge to real-world situations through participation in social activities.

[0103] Classroom participation characteristics include head-up rate, interaction response frequency, and facial focus index. Head-up rate is the ratio of the number of student head-up images to the total number of images within a preset time period, reflecting the student's level of focus in class. Interaction response frequency is the ratio of the number of effective verbal responses recognized from students in class to the total class time, reflecting the student's activity level in classroom discussions. The facial focus index is calculated using heart rate variability data and reflects the student's emotional focus. Learning self-discipline characteristics include attendance rate, on-time homework submission rate, and time spent learning from extracurricular resources.

[0104] Attendance rate refers to the ratio of actual attendance to the required attendance, reflecting students' learning discipline. On-time assignment submission rate refers to the ratio of the number of assignments submitted on time to the total number of assignments submitted, reflecting students' sense of responsibility towards their studies. Extracurricular resource learning time refers to the total time students spend using the ideological and political education digital resource platform outside of class, reflecting students' learning initiative. The social practice dimension includes the number of volunteer service attendance checks and the percentage of effective participation in group collaboration projects. Volunteer service attendance checks refer to the number of times students participate in volunteer service activities, reflecting their sense of social responsibility. The percentage of effective participation in group collaboration projects refers to the ratio of the number of effective participations in group projects to the total number of participations, reflecting students' teamwork and practical abilities.

[0105] Preferably, when constructing a multi-dimensional feature vector for the effectiveness of ideological and political education, the head-up rate can be calculated by analyzing facial image data collected by a camera. Specific steps include identifying facial features in the images, determining whether students are looking up, and calculating the ratio of the number of images with head-up to the total number of images. The interaction response frequency can be calculated by analyzing voice data collected by a microphone array. Specific steps include speech recognition and keyword extraction, counting the number of effective voice responses from students in class, and calculating the ratio with the total class time. The facial focus index can be calculated by analyzing heart rate variability data collected by a smart bracelet. Specific steps include calculating the difference between the heart rate variability at each time point and the baseline heart rate variability at rest, and obtaining the focus index through time series analysis. These detailed calculation steps ensure the accuracy and reliability of the feature vector, providing more precise data support for subsequent evaluation of the effectiveness of ideological and political education.

[0106] In some embodiments, in step S301, the head-up rate, the interaction response frequency, and the facial focus index are calculated as follows:

[0107] The head-up rate From the formula Calculations show that

[0108] in, This indicates the number of student head-up image frames captured within a preset time period. This represents the total number of student image frames collected within the preset time period;

[0109] The interactive response frequency R is given by the formula Calculations show that

[0110] in, This indicates the number of valid verbal responses that were recognized from a student during class. Indicates the total class time;

[0111] The facial focus index (F) is calculated using the formula...

[0112]

[0113] The calculation shows that, Indicates the total number of sampling periods. This represents the heart rate variability data at time t. This represents the student's baseline heart rate variability data at rest.

[0114] The head-up rate is calculated by analyzing facial image data captured by the camera. Within a preset time period, the system counts the number of frames showing students looking up, i.e., the ratio of the number of frames where the student's face is facing the camera and their eyes are open, to the total number of frames. The interaction response frequency is calculated by analyzing voice data captured by the microphone array. The system counts the number of effective voice responses recognized by students in class, i.e., the ratio of the number of times a student speaks and the content is relevant to class discussion, to the total class time. The facial focus index is calculated by analyzing heart rate variability data collected by a smart bracelet. The system calculates the difference between the heart rate variability at each time point and the student's baseline heart rate variability at rest, and derives the focus index through time series analysis. These parameters are set based on the actual collected data and preset time periods, and these methods quantify student classroom participation.

[0115] Preferably, when calculating the head-up rate, deep learning algorithms can be used to analyze the image, identify students' facial features, and determine whether they are looking up. When calculating the interaction response frequency, speech recognition technology can be used to convert students' speech data into text, and natural language processing technology can be used to extract keywords and semantic information, counting the number of effective voice responses from students in the classroom. When calculating the facial focus index, signal processing technology can be used to denoise and extract features from heart rate variability data, combined with students' resting state baseline data, to calculate the difference in heart rate variability at each time point, and then derive the focus index through time series analysis. These detailed calculation steps ensure the accuracy and reliability of feature calculations, providing more precise data support for subsequent evaluation of ideological and political education effectiveness.

[0116] In some embodiments, in step S4, the pre-trained ideological and political education effectiveness evaluation model is obtained through the following steps:

[0117] S401: Collect feature vectors of the ideological and political education effects of multiple students within a historical time period as a training sample set;

[0118] S402: Obtain the real effect score and real effect level corresponding to the training sample set, as evaluated by ideological and political education experts, and use them as training labels;

[0119] S403: Using the training sample set as input and the training labels as expected output, supervise the training of the machine learning model until the model loss function converges to obtain the ideological and political education effect evaluation model.

[0120] In implementation, firstly, feature vectors of ideological and political education effectiveness are collected from multiple students over a historical period. These feature vectors contain data on students' participation in class, self-discipline in learning, and social practice. Secondly, the actual effectiveness scores and levels, assessed by ideological and political education experts, corresponding to these feature vectors are obtained. This data serves as training labels to guide the model's learning process. Finally, using the feature vectors as input and the actual effectiveness scores and levels as expected outputs, the machine learning model undergoes supervised training until the model's loss function converges, resulting in the final evaluation model. In this process, the machine learning model learns from a large amount of sample data, automatically identifying key features affecting the effectiveness of ideological and political education and establishing corresponding evaluation rules.

[0121] Preferably, when collecting historical data, a specific time range and sample size can be set to ensure data diversity and representativeness. When selecting a machine learning model, a gradient boosting decision tree model can be used. This model minimizes the loss function by iteratively adding decision trees, exhibiting good predictive performance and interpretability. During training, parameters such as the learning rate, tree depth, and number can be set to optimize model performance. Furthermore, methods such as cross-validation can be used to evaluate the model's generalization ability, ensuring its stability and accuracy on different datasets.

[0122] In some embodiments, in step S403, the machine learning model is a gradient boosting decision tree model; the gradient boosting decision tree model minimizes a loss function by iteratively adding decision trees, and the model output of the k-th iteration is... Represented as:

[0123]

[0124] in, This represents the predicted value of the i-th sample after the k-th iteration. This represents the predicted value of the i-th sample after the (k-1)-th iteration. The learning rate is used to control the contribution of each decision tree. This represents the decision tree model added in the k-th iteration. Let represent the feature vector of the ideological and political education effect of the i-th sample;

[0125] The loss function is the mean squared error function, with a regularization term added. Its formula is:

[0126]

[0127] in, This represents the value of the loss function. This indicates the number of samples in the training sample set. This represents the true performance score of the i-th sample. This represents the model's prediction score for the i-th sample. This represents the total number of leaf nodes in the model. This represents the weight value of the j-th leaf node. and This represents the regularization parameter that controls the complexity of the model.

[0128] In this invention, the gradient boosting decision tree model is a machine learning algorithm that iteratively optimizes the model's predictive performance by adding decision trees. The core of this model lies in minimizing a loss function while preventing overfitting through regularization. This method can effectively handle complex nonlinear relationships and automatically select important features, thus providing an accurate and reliable model for the quantitative evaluation of ideological and political education effectiveness.

[0129] Specifically, the model iteratively builds multiple decision trees. In each iteration, the model calculates the residual between the current prediction and the true value, and trains a new decision tree based on these residuals to correct previous prediction errors. (Learning rate) The _________ is an important parameter that controls the contribution of each decision tree to the final prediction result, typically taking a value between 0 and 1. The number of decision trees, i.e., the number of iterations, is also a key parameter, determining the model's complexity and fitting ability. Furthermore, the regularization term controls the total number T of leaf nodes in the model and the weights of the leaf nodes. This further prevents model overfitting. Regularization parameters and Used to balance model complexity and prediction error.

[0130] Preferably, the operational steps can be further refined when constructing and training the gradient boosting decision tree model. For example, during the model initialization phase, a small learning rate, such as 0.1, can be set to ensure stable convergence of the model. The number of decision trees can be adjusted based on the results of cross-validation, typically between 100 and 500. Regularization parameters and Optimization can be achieved through methods such as grid search to find the best combination of parameters. In each iteration, the model calculates the predicted value for each sample and updates the model parameters based on the mean squared error loss function. In this way, the model can gradually learn the complex relationship between student characteristics and the effectiveness of ideological and political education, ultimately outputting a quantitative comprehensive evaluation score. This method not only improves the accuracy of the evaluation but also provides educators with intuitive evaluation results and feedback, thereby better guiding the improvement of ideological and political education.

[0131] In some embodiments, the qualitative evaluation label is obtained by mapping the comprehensive evaluation score; the mapping rule is as follows: when the comprehensive evaluation score is in the range [90, 100], the qualitative evaluation label is excellent; when the comprehensive evaluation score is in the range [75, 90), the qualitative evaluation label is good; when the comprehensive evaluation score is in the range [60, 75), the qualitative evaluation label is acceptable; when the comprehensive evaluation score is in the range [0, 60), the qualitative evaluation label is to be improved.

[0132] In this invention, step S4 involves mapping the comprehensive assessment score to qualitative assessment labels. This process transforms the quantitative assessment results into more intuitive and easily understood qualitative descriptions, enabling educators and students to quickly grasp the overall level of effectiveness in ideological and political education. The qualitative assessment labels are divided according to different ranges of the comprehensive assessment score, such as excellent, good, satisfactory, and needing improvement, clearly reflecting students' performance in ideological and political education. This mapping method not only improves the readability of the assessment results but also provides clear guidance for subsequent educational decisions.

[0133] Specifically, the mapping rules for qualitative assessment labels are based on the intervals defined by the comprehensive assessment scores. When the comprehensive assessment score falls within the interval [90, 100], the qualitative assessment label is "Excellent," indicating that the student performed exceptionally well in ideological and political education, achieving a high level in all aspects. When the comprehensive assessment score falls within the interval [75, 90), the qualitative assessment label is "Good," indicating that the student performed well but still has room for improvement. When the comprehensive assessment score falls within the interval [60, 75), the qualitative assessment label is "Pass," indicating that the student met the basic educational requirements. When the comprehensive assessment score falls within the interval [0, 60), the qualitative assessment label is "Needs Improvement," indicating that the student has many problems in ideological and political education and requires further education and guidance. These intervals and labels are determined based on educational goals and actual needs, aiming to provide educators with clear feedback and guidance.

[0134] After calculating the comprehensive assessment score, the system automatically maps labels according to preset interval rules. To ensure the accuracy of the mapping, a mapping table can be set up to associate each score interval with a corresponding qualitative assessment label. In practical applications, the system will query the mapping table based on the student's comprehensive assessment score and quickly output the corresponding qualitative assessment label. Furthermore, to improve the flexibility of the assessment, the division of score intervals and the definition of labels can be adjusted according to different educational scenarios and needs. For example, in some cases, the labels for improvement can be further subdivided to more accurately reflect the student's problems. In this way, qualitative assessment labels not only provide intuitive assessment results but also provide a basis for educators to develop personalized educational improvement plans.

[0135] In some embodiments, step S5, generating a personalized report of suggestions for improving ideological and political education, specifically includes:

[0136] S501: Compare the scores of each dimension of the ideological and political education effect feature vector with a preset threshold to identify the weak dimensions that are below the threshold.

[0137] S502: Based on a predefined knowledge base of improvement strategies, match at least one improvement suggestion for each of the aforementioned weakness dimensions;

[0138] S503: Summarize and structure all the matched improvement suggestions to generate the personalized ideological and political education improvement suggestion report.

[0139] The process of generating a personalized improvement suggestion report includes the following key steps: First, the scores of each dimension of the ideological and political education effectiveness feature vector are compared with preset thresholds to identify weak dimensions that fall below the threshold. These dimensions include classroom participation, learning self-discipline, and social practice. For example, if a student's classroom participation score is below the preset threshold, it indicates a deficiency in classroom participation. Second, based on a predefined improvement strategy knowledge base, at least one improvement suggestion is matched for each weak dimension. The improvement strategy knowledge base is a collection of various teaching improvement methods, each targeting a specific weak dimension. Finally, all matched improvement suggestions are summarized and structured to generate a personalized improvement suggestion report. This report will provide educators and students with clear directions and specific measures for improvement.

[0140] Preferably, the operational steps can be further refined when generating personalized improvement suggestion reports. For example, when identifying weaknesses, specific thresholds can be set, such as 70 points for classroom participation, 65 points for self-discipline, and 60 points for social practice. For dimensions below these thresholds, the system will automatically match corresponding improvement suggestions from the improvement strategy knowledge base. The improvement strategy knowledge base can contain various specific improvement measures, such as increasing classroom interaction, setting up learning supervision mechanisms, and organizing social practice projects. When generating reports, the system can sort the dimensions according to their importance, prioritizing the display of improvement suggestions for key dimensions. Furthermore, the reports can be presented in the form of tables or charts to help users understand the improvement suggestions more intuitively. In this way, personalized improvement suggestion reports not only provide clear directions for improvement but also offer educators and students specific and actionable improvement measures, thereby effectively improving the quality and relevance of ideological and political education.

[0141] In some embodiments, in step S4, the ideological and political education effectiveness evaluation model is a long short-term memory network model based on an attention mechanism; the process by which the long short-term memory network model based on an attention mechanism processes the time-seriesd feature vector of the ideological and political education effectiveness includes:

[0142] S1001: Feature vector of ideological and political education effectiveness arranged in time series Input the data into a Long Short-Term Memory (LSTM) network layer to obtain the hidden state at each time step. ;

[0143] in, This represents the input feature vector sequence. This represents the feature vector at time step t. Represents the hidden state sequence, This represents the hidden state at time step t.

[0144] S1002: Calculate the attention weights of the hidden state at each time step through the attention layer. The calculation formula is as follows:

[0145]

[0146] in, This represents the attention weight at the i-th time step. It is an alignment model function. , , This is the trainable parameter matrix for the attention mechanism. This represents the hidden state at the i-th time step. This represents the hidden state at the last time step;

[0147] S1003: Perform a weighted summation on the hidden state sequence to obtain the context vector.

[0148] in, This represents a context vector that encapsulates important information from all time steps.

[0149] S1004: Pass the context vector c through a fully connected output layer to obtain the final comprehensive evaluation score.

[0150] The processing steps of the attention-based Long Short-Term Memory (LSTM) network model include the following: First, the feature vectors of ideological and political education effectiveness, arranged in a time sequence, are input into the LSTM network layer to obtain the hidden state at each time step. The LSTM layer processes the input feature vector sequence, captures the temporal dependencies within the sequence, and outputs the hidden state at each time step. These hidden states contain important information from each time point in the sequence. Next, the attention layer calculates the attention weights for the hidden states at each time step. The attention mechanism dynamically allocates weights based on the hidden states of the current time step and other time steps, thereby highlighting important time steps.

[0151] The hidden state sequence is weighted and summed to obtain a context vector, which encapsulates the key information from all time steps. This context vector is then passed through a fully connected output layer to obtain the final comprehensive evaluation score. In this process, the model learns patterns and relationships in the data, automatically identifying the features and time points that have the greatest impact on the effectiveness of ideological and political education for students.

[0152] When inputting data, the feature vectors of ideological and political education effectiveness can be arranged in chronological order to ensure the model can capture dynamic changes in the time series. In implementing the attention mechanism, an alignment model function can be defined to calculate attention weights; this function typically involves a similarity measure between hidden states, such as a dot product or weighted sum. During model training, appropriate hyperparameters, such as the number of units in the LSTM layer and the dimension of the attention layer, can be set to optimize model performance. Furthermore, optimization algorithms such as backpropagation and gradient descent can be used to train the model, enabling it to minimize the error between predicted and true scores. After training, the model's performance can be evaluated using a test set and adjusted as needed. In this way, the attention-based Long Short-Term Memory (LSTM) network model can effectively process time-series data on ideological and political education effectiveness, providing educators with accurate evaluation results.

[0153] The above embodiments of the present invention have the following beneficial effects:

[0154] 1. By deploying various IoT sensors in ideological and political education settings, multimodal behavioral data of students is collected, enabling comprehensive and real-time monitoring of student behavior. This method not only captures students' classroom performance, such as facial expressions, body movements, and voice interactions, but also records physiological data such as heart rate variability and physical activity intensity, as well as extracurricular learning behavior and attendance. This provides a rich and multidimensional data foundation for evaluating the effectiveness of ideological and political education, solving the problems of single data sources and limited dimensions in traditional evaluation methods, making the evaluation results more comprehensive and objective.

[0155] 2. Data preprocessing and fusion techniques are employed to clean, timestamp-align, and normalize the collected multimodal behavioral data, which is then integrated according to student identification identifiers to generate a standardized student behavior dataset. This process effectively eliminates outlier and noisy data, unifies the data's time reference and numerical units, and improves data quality and usability. This provides high-quality data support for subsequent feature vector construction and evaluation model training, solving the problem of inaccurate evaluation results caused by non-standard data processing and poor data quality in existing technologies, and ensuring the stability and reliability of the evaluation model.

[0156] 3. A multi-dimensional feature vector of ideological and political education effectiveness is constructed, and a pre-trained assessment model is used to output quantitative scores and qualitative labels, while simultaneously generating personalized improvement suggestion reports. This method not only quantifies the effectiveness of ideological and political education but also visually displays the assessment results through a graphical interface, helping educators quickly understand students' learning status and problems. Furthermore, the personalized improvement suggestion reports provide teachers with targeted directions for teaching improvement, addressing the lack of feedback mechanisms and personalized guidance in traditional assessment methods, and improving the relevance and effectiveness of ideological and political education.

[0157] The above description is merely an explanation of some preferred embodiments of the present invention and the technical principles employed. Those skilled in the art should understand that the scope of the invention as described in the embodiments of the present invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A method for evaluating the effectiveness of ideological and political education based on the Internet of Things, characterized in that, Includes the following steps: S1: Collect students' multimodal behavior data through multiple IoT sensing devices deployed in ideological and political education scenarios, and upload the multimodal behavior data to the central processing server; S2: The central processing server preprocesses and fuses the received multimodal behavior data to generate a standardized student behavior dataset; S3: Based on the standardized student behavior dataset, construct a multi-dimensional feature vector of ideological and political education effectiveness, wherein the feature vector of ideological and political education effectiveness includes classroom participation dimension features, learning self-discipline dimension features and social practice dimension features; S4: Input the feature vector of the ideological and political education effect into a pre-trained ideological and political education effect evaluation model, and the ideological and political education effect evaluation model outputs a quantitative comprehensive evaluation score and a qualitative evaluation label. S5: Display the comprehensive evaluation score and the qualitative evaluation label through a visual interface, and generate a personalized report of suggestions for improving ideological and political education.

2. The method as described in claim 1, characterized in that, Step S1 specifically includes: S101: Collect students' facial image data and body movement video data in the classroom through the camera in the IoT sensing device; S102: Collect students' classroom voice interaction data through the microphone array in the IoT sensing device; S103: Collect students' heart rate variability data and physical activity intensity data through the smart bracelet in the IoT sensing device; S104: Collect student classroom attendance time data through the access card reader in the IoT sensing device; S105: Collect log data of students accessing the ideological and political education digital resource platform through the students' personal smart terminal devices.

3. The method as described in claim 1, characterized in that, The data preprocessing and fusion described in step S2 specifically includes: S201: Perform data cleaning on the multimodal behavior data to remove abnormal and noisy data; S202: Perform timestamp alignment on the cleaned multimodal behavioral data to unify all data into the same time coordinate system; S203: Perform data normalization processing on the timestamp-aligned multimodal behavior data to make all data have the same numerical dimension; S204: The normalized multimodal behavior data is associated and integrated according to student identity identifiers to generate the standardized student behavior dataset for each student.

4. The method as described in claim 1, characterized in that, In step S3, constructing a multi-dimensional feature vector of ideological and political education effectiveness specifically includes: S301: Based on the standardized student behavior dataset, calculate the classroom participation dimension features, which include head-up rate, interaction response frequency, and facial focus index. S302: Based on the standardized student behavior dataset, calculate the learning self-discipline dimension features, which include attendance rate, on-time homework submission rate, and extracurricular resource learning time. S303: Based on the standardized student behavior dataset, calculate the social practice dimension features, which include the number of volunteer service check-ins and the percentage of effective speeches in group collaboration projects.

5. The method as described in claim 4, characterized in that, In step S301, the head-up rate, the interaction response frequency, and the facial focus index are calculated as follows: The head-up rate ( From the formula Calculations show that in, This indicates the number of student head-up image frames captured within a preset time period. This represents the total number of student image frames collected within the preset time period; The interaction response frequency (R) is given by the formula Calculations show that in, This indicates the number of valid verbal responses that were recognized from a student during class. Indicates the total class time; The facial focus index (F) is calculated using the formula... The calculation shows that, Indicates the total number of sampling periods. This represents the heart rate variability data at time t. This represents the student's baseline heart rate variability data at rest.

6. The method as described in claim 1, characterized in that, In step S4, the pre-trained ideological and political education effectiveness evaluation model is obtained through the following steps: S401: Collect feature vectors of the ideological and political education effects of multiple students within a historical time period as a training sample set; S402: Obtain the real effect score and real effect level corresponding to the training sample set, as evaluated by ideological and political education experts, and use them as training labels; S403: Using the training sample set as input and the training labels as expected output, supervise the training of the machine learning model until the model loss function converges to obtain the ideological and political education effect evaluation model.

7. The method as described in claim 6, characterized in that, In step S403, the machine learning model is a gradient boosting decision tree model; the gradient boosting decision tree model minimizes a loss function by iteratively adding decision trees, and the model output of the k-th iteration is... Represented as: in, This represents the predicted value of the i-th sample after the k-th iteration. This represents the predicted value of the i-th sample after the (k-1)-th iteration. The learning rate is used to control the contribution of each decision tree. This represents the decision tree model added in the k-th iteration. Let represent the feature vector of the ideological and political education effect of the i-th sample; The loss function is the mean squared error function, with a regularization term added. Its formula is: in, This represents the value of the loss function. This indicates the number of samples in the training sample set. This represents the true performance score of the i-th sample. This represents the model's prediction score for the i-th sample. This represents the total number of leaf nodes in the model. This represents the weight value of the j-th leaf node. and This represents the regularization parameter that controls the complexity of the model.

8. The method as described in claim 1, characterized in that, In step S4, the qualitative evaluation label is obtained by mapping the comprehensive evaluation score; the mapping rule is as follows: when the comprehensive evaluation score is in the interval [90, 100], the qualitative evaluation label is "excellent"; when the comprehensive evaluation score is in the interval [75, 90), the qualitative evaluation label is "good"; when the comprehensive evaluation score is in the interval [60, 75), the qualitative evaluation label is "passable"; when the comprehensive evaluation score is in the interval [0, 60), the qualitative evaluation label is "needs improvement".

9. The method as described in claim 1, characterized in that, Step S5, which involves generating a personalized report of suggestions for improving ideological and political education, specifically includes: S501: Compare the scores of each dimension of the ideological and political education effect feature vector with a preset threshold to identify the weak dimensions that are below the threshold. S502: Based on a predefined knowledge base of improvement strategies, match at least one improvement suggestion for each of the aforementioned weakness dimensions; S503: Summarize and structure all the matched improvement suggestions to generate the personalized ideological and political education improvement suggestion report.

10. The method as described in claim 1, characterized in that, In step S4, the ideological and political education effectiveness evaluation model is a long short-term memory network model based on the attention mechanism; the process by which the long short-term memory network model based on the attention mechanism processes the time-seriesd feature vector of the ideological and political education effectiveness includes: S1001: Feature vector of ideological and political education effectiveness arranged in time series Input the data into a Long Short-Term Memory (LSTM) network layer to obtain the hidden state at each time step. ; in, This represents the input feature vector sequence. This represents the feature vector at time step t. Represents the hidden state sequence, This represents the hidden state at time step t. S1002: Calculate the attention weights of the hidden state at each time step through the attention layer. The calculation formula is as follows: in, This represents the attention weight at the i-th time step. It is an alignment model function. , , This is the trainable parameter matrix for the attention mechanism. This represents the hidden state at the i-th time step. This represents the hidden state at the last time step; S1003: Perform a weighted summation on the hidden state sequence to obtain the context vector. in, This represents a context vector that encapsulates important information from all time steps. S1004: Pass the context vector c through a fully connected output layer to obtain the final comprehensive evaluation score.

Images