A method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching

Abstract

The present invention relates to the field of teaching technology, and discloses a method for analyzing learning conditions and predicting learning effects of online and offline hybrid teaching, comprising the following steps: Step 1, collecting students' online learning data, offline classroom data, emotional data and environmental data; Step 2, preprocessing the collected multi-source data, removing outliers and performing standardization processing; Step 3, fusing multimodal data to extract students' learning behaviors, emotional fluctuations, academic performance and other characteristics. Personalized learning support is achieved through comprehensive analysis of multi-dimensional data such as students' online learning behaviors, offline classroom performances and programming tasks. Teachers can formulate personalized learning plans and tutoring strategies for each student based on the learning progress, ability analysis and emotional fluctuation prediction results provided by the system, to ensure that each student obtains the most appropriate teaching support within their ability, thereby improving learning effects.

Description

Technical Field

[0001] The present invention relates to the field of teaching technology, and specifically to a method for analyzing learning situations and predicting learning effects of online and offline hybrid teaching. Background Art

[0002] With the development of information technology, online education has developed rapidly around the world, and online learning has become the main teaching method for many educational institutions. However, traditional online education platforms still face many challenges in teaching effect evaluation and learning situation analysis, especially in how to achieve accurate learning effect prediction and personalized teaching support. Existing online teaching systems mainly rely on students' learning behavior data (such as video viewing time, answering scores, etc.) for analysis, but these data can often only reflect part of the students' learning situation, lack of comprehensive analysis of students' learning process, emotional changes and interactive behavior, resulting in low accuracy of learning situation evaluation and learning effect prediction results.

[0003] At the same time, offline classroom teaching, as a traditional teaching mode, has the advantages of strong interactivity and teachers can understand students' learning status in real time, but its teaching resources and time are limited, and teachers cannot pay full attention to the personalized learning needs of each student. In traditional classroom teaching, teachers usually evaluate students' learning effects through face-to-face communication and written tests, but this evaluation method is often lagging and cannot reflect students' learning status in a timely manner. It also ignores the influence of non-cognitive factors such as students' emotional fluctuations, initiative and participation in the learning process.

[0004] In order to make up for the shortcomings of traditional online education and offline classroom teaching, many educational technology researchers and institutions have begun to explore online and offline hybrid teaching models in recent years. Hybrid teaching aims to improve learning outcomes, increase learning participation, and optimize the allocation of teaching resources by combining the flexibility of online teaching with the interactivity of offline teaching. However, although the hybrid teaching model has gradually become mainstream, how to scientifically and accurately analyze learning situations, predict learning outcomes, and provide personalized learning support is still the core problem facing this model.

[0005] At present, most online learning platforms still lack comprehensive analysis tools for students' comprehensive learning situations, cannot effectively integrate online and offline data, and lack dynamic evaluation mechanisms for multi-dimensional information such as students' emotional states and interactive behaviors. In addition, existing learning effect prediction methods usually rely too much on static performance data or behavioral data, ignoring non-cognitive factors such as students' emotional changes, participation, and learning motivation during the learning process, which leads to large errors in the prediction results and affects the accuracy and effectiveness of teaching decisions.

[0006] Therefore, how to achieve accurate learning situation analysis and learning effect prediction in blended teaching, especially dynamic analysis based on multi-dimensional data, is still a technical problem that needs to be solved urgently in the field of educational technology. Summary of the invention

[0007] In view of the shortcomings of the prior art, the present invention provides a method for analyzing learning situation and predicting learning effect in online and offline hybrid teaching, which solves the problem of how to achieve accurate learning situation analysis and learning effect prediction through multi-dimensional data fusion and analysis in online and offline hybrid teaching, thereby providing personalized teaching support and optimizing teaching decisions.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching, comprising the following steps: Step 1: Collect students’ online learning data, offline classroom data, emotional data, and environmental data; Step 2: Preprocess the collected multi-source data, remove outliers and perform standardization; Step 3: Fuse multimodal data to extract characteristics of students’ learning behaviors, emotional fluctuations, and academic performance; Step 4: Construct a student's learning profile, including academic ability profile, learning behavior profile, emotional profile, and learning resource preference profile; Step 5: Based on the learning profile and student historical data, predict learning outcomes and provide real-time feedback and intervention strategies.

[0009] Preferably, the pretreatment step comprises: De-noise the learning behavior data to remove invalid information; Use sliding window method to process time series data; The sentiment dictionary analysis method is used to extract the emotional fluctuation characteristics of the emotional data; Academic performance data were standardized to eliminate dimension effects.

[0010] Preferably, the multimodal data fusion step includes: Jointly model learning behavior, emotional data, and academic performance data using cross-attention mechanisms; Different deep learning models are used for each data modality, including LSTM for learning behavioral data, BiLSTM for sentiment data, and XGBoost for academic performance data.

[0011] Preferably, the learning portrait construction step includes: Build academic ability portraits through knowledge tracking models and update students’ knowledge points in real time; Use clustering algorithms to classify learning behaviors, identify students’ learning styles, and generate learning behavior portraits based on the styles; Analyze students’ emotional fluctuations based on emotional data and generate emotional portraits; Generate a learning resource preference profile based on learning history and identify the types of learning resources preferred by students.

[0012] Preferably, the learning effect prediction step includes: Use the Transformer model to predict students’ learning progress and capture long-term dependencies; Based on the multi-task learning model, multiple tasks such as academic performance, emotional fluctuations, and learning attitude are predicted simultaneously; Weighted voting and Stacking models are used to fuse the prediction results of different data sources.

[0013] Preferably, the real-time feedback step includes: Adjust the difficulty of learning tasks in real time according to students’ learning progress and emotional fluctuations; Provide personalized learning resources and learning task recommendations for students who are lagging behind in their studies; Generate encouraging feedback based on students' emotional states.

[0014] Preferably, the intervention strategy includes: Providing targeted remedial materials and exercises to students with poor academic performance; Recommend emotion management techniques and psychological counseling resources to students with greater emotional fluctuations; Students with passive learning styles are motivated to learn through motivational content and interactive design.

[0015] Preferably, the personalized learning path recommendation step includes: Dynamically recommend appropriate learning paths based on students’ learning profiles and learning styles; According to changes in learning behavior data and emotional portraits, the learning path is adjusted in real time, and customized learning resources are pushed.

[0016] Preferably, the learning resource recommendation step includes: Based on students’ learning history data, collaborative filtering algorithms are used to recommend learning resources for students with similar learning styles. Dynamically adjust resource recommendations based on students’ emotional portraits.

[0017] Preferably, the real-time feedback and intervention steps are implemented through a learning management system, which includes a learning data acquisition module, a data preprocessing module, a learning effect prediction module, a real-time feedback module and a personalized recommendation module. The modules interact in real time through the network to form a closed-loop feedback mechanism.

[0018] The present invention provides a method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching. It has the following beneficial effects: 1. The present invention realizes personalized learning support through comprehensive analysis of multi-dimensional data such as students' online learning behavior, offline classroom performance, and programming tasks. Teachers can formulate personalized learning plans and tutoring strategies for each student based on the learning progress, ability analysis, and emotional fluctuation prediction results provided by the system to ensure that each student obtains the most appropriate teaching support within their ability, thereby improving learning outcomes.

[0019] 2. The present invention combines machine learning models with students' behavioral data, programming ability and emotional feedback to accurately predict students' academic performance and ability development trends. This prediction can help teachers discover learning bottlenecks and potential problems in advance, adjust teaching strategies in a timely manner, and ensure that each student can receive effective guidance and help in the teaching process, thereby improving the overall learning effect.

[0020] 3. The present invention can dynamically adjust teaching strategies by real-time tracking of students' learning process, emotional changes and performance. When the system identifies that students have obvious deficiencies in a certain area, it will automatically make corresponding teaching optimization suggestions to help teachers flexibly adjust course content, teaching methods and tutoring strategies to optimize classroom teaching effects.

[0021] 4. The present invention not only focuses on students' learning results, but also on students' participation and emotional state throughout the learning process. By analyzing students' interactive behaviors, emotional feedback, etc., teachers can timely identify students' emotional fluctuations and participation problems, promote students' active participation in the learning process, reduce learning anxiety, increase learning interest, and thus improve learning motivation and learning efficiency.

[0022] 5. Through comprehensive and systematic data analysis, the present invention provides teachers with in-depth learning situation analysis reports, helping teachers to extract key issues and advantages and disadvantages of students' learning from the data. This enables teachers to make teaching decisions more accurately and formulate teaching content and schedule in a targeted manner, thereby improving teaching effectiveness and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the specification 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.

[0025] Please see attached Figure 1 The embodiment of the present invention provides a method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching, including: Step 1: Data collection and preprocessing Step S1 is the basic link of the present invention, which aims to comprehensively collect information such as students' online learning behavior, offline classroom performance, emotional fluctuations and environmental factors through multi-channel and multi-modal data collection technology, and pre-process the collected data. Through the implementation of this step, the accuracy and comprehensiveness of subsequent multimodal data fusion and learning effect prediction can be ensured. The core of this step is the integrity, validity and availability of the data. Combining the characteristics of different data sources, a specific collection and processing method is adopted to provide high-quality data input for subsequent steps.

[0026] 1. Data Collection In this embodiment, in view of the characteristics of online and offline hybrid teaching, student data is collected from the following dimensions: Online learning data collection: In this embodiment, the online learning behavior of students is recorded through the online learning platform (such as Moodle or teaching management system) used by the students, including login time, learning time, video viewing progress, page dwell time and test submission status.

[0027] Students' homework scores and online test scores are collected in real time through the platform interface, and include detailed records such as answering time, error distribution, and question scores.

[0028] Students’ interactive behavior data are collected through online discussion boards and question-and-answer systems, including the number of posts, questions, and likes.

[0029] Offline classroom data collection: In this embodiment, classroom interactive equipment (such as infrared sensors, student feedback devices) is used to record students' behavioral data in the classroom, such as the number of times they raise their hands, the number of times they answer questions in class, and the interaction with classmates.

[0030] In response to the actual needs of majors such as stomatology, students' experimental performance is recorded through experimental operation equipment, including operation trajectory, experimental duration, success rate and teacher scores.

[0031] Emotional data collection: In this embodiment, facial expressions and voice features of students during the learning process are obtained through cameras and voice acquisition devices, and expression recognition technology (such as OpenCV combined with deep learning models) and voice emotion analysis technology (such as Mel spectrum combined with CNN models) are combined to extract students' emotional fluctuation information.

[0032] In addition, information on students’ psychological status (such as anxiety, learning motivation, satisfaction, etc.) was collected through questionnaire survey tools. The design of the questionnaire was based on commonly used psychological scales (such as GAD-7, PSQI, etc.).

[0033] Environmental data collection: In this embodiment, the online learning platform collects students' device usage (such as device type, operating system version) and network status (such as bandwidth, latency, etc.). This information is automatically collected through network logs and API interfaces.

[0034] Regarding the classroom environment, the noise level and equipment operation status of the classroom are collected to evaluate the potential impact of the external environment on learning outcomes.

[0035] 2. Data preprocessing In this embodiment, a unified preprocessing operation is performed on the collected multi-source data to ensure the quality and consistency of the data. The specific steps are as follows: Data cleaning: Remove invalid data: Filter abnormal behavior data (such as records with too short or too long learning time) by setting rules. Outlier detection uses the IQR (interquartile range) method to automatically remove extreme values.

[0036] Processing duplicate data: Merge duplicated behavior entries in the platform log to ensure data uniqueness.

[0037] Data completion: For missing learning behavior data (such as a student not completing a test record in a certain stage), this embodiment uses the KNN (K nearest neighbor) interpolation method to fill in the missing values ​​to avoid data analysis failure due to missing values.

[0038] For some unrecorded classroom interaction data, the missing content is inferred through the time series correlation of the surrounding data.

[0039] Standardization and normalization: In this embodiment, all numerical data (such as homework scores, study time, experiment scores, etc.) are subjected to Min-Max normalization processing to map the data values ​​to the range of 0-1.

[0040] One-hot encoding is performed on categorized data (such as classroom performance ratings) to facilitate subsequent feature analysis.

[0041] Timing processing: For behavioral data (such as learning time, video viewing history) and emotional data (such as emotional change curves), the sliding window method is used to divide the data sequence into fixed-length time segments to construct input samples that adapt to the time series model. The sliding window length is set to 5 minutes based on the data density and learning stage characteristics.

[0042] Feature Dimensionality Reduction: In this embodiment, principal component analysis (PCA) technology is used to reduce the dimension of multi-dimensional behavior data and remove redundant information. For highly correlated features (such as video viewing time and page dwell time), the most discriminative principal component is retained.

[0043] Through the above implementation steps, step S1 realizes the processing chain from multi-source data collection to high-quality data input, ensuring the integrity, reliability and consistency of the collected data, and providing a solid foundation for subsequent multimodal data fusion and learning effect prediction.

[0044] Step 2: Multimodal data fusion and feature extraction The purpose of step S2 is to fuse and analyze the collected multi-dimensional data and establish a unified learning behavior data model. The key to this step is to use scientific data fusion methods to integrate various information such as online and offline learning behaviors, emotional fluctuations and environmental factors to form a more comprehensive learning situation analysis model. This process can extract valuable learning behavior characteristics and provide solid data support for subsequent learning effect predictions. The focus of this step is on the multimodal fusion of data to ensure that data from different sources can be effectively combined to show the overall picture of student learning.

[0045] 1. Data fusion framework design In this embodiment, a multimodal data fusion method is used to integrate different types of data (such as online learning behavior data, offline classroom behavior data, emotional data, and environmental data). The data fusion framework is divided into the following main parts: Multimodal Data Integration: In this embodiment, an integrated model based on feature fusion is designed for online learning data, offline classroom data, emotional data and environmental data. All raw data are first converted into a standardized format (such as timestamps, behavior labels, emotional scores, etc.), and then cleaned and processed by the data preprocessing module. Subsequently, the features of each data source are concatenated to form a unified feature vector as input. For time series data (such as learning duration and emotional fluctuations), the sliding window method is used for segmented processing, and the feature vectors of each segment of data are merged into a unified time series feature.

[0046] Feature selection and dimensionality reduction: To avoid data redundancy and overfitting, PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) are combined for feature dimensionality reduction and selection. PCA is used to reduce feature dimensions and retain the principal components in the data; LDA helps find the features that can best distinguish the learning status of different students in high-dimensional data. This method can maximize the retention of valuable features and improve the learning efficiency and prediction accuracy of the model.

[0047] 2. Student learning status modeling Based on data fusion, this embodiment further models the students' learning status and uses a mixed model (such as a Gaussian mixture model, GMM) to classify the students' learning status. According to the students' online behavior, emotional state and environmental factors, a multi-dimensional learning status model is constructed, which is divided into different learning status categories, such as "efficient learning status", "inefficient learning status", "emotional fluctuation status", etc.

[0048] Classification of learning status: In this embodiment, a clustering algorithm (such as a K-means algorithm) is used to perform unsupervised learning on the multimodal fused data to obtain the student's learning state category. The most representative learning state is automatically identified through the behavioral characteristics of different students in the training data. These states include but are not limited to learning progress, emotional stability, and the impact of the external environment on learning. Each learning state category will include a set of representative behavioral characteristics, such as "concentrated learning state" may be manifested as long learning time, high homework scores, and low emotional fluctuations.

[0049] Dynamic adjustment of learning status: Combined with students' long-term learning trajectory and emotional fluctuations, a dynamic learning state adjustment mechanism is adopted. Through time series analysis (such as HMM hidden Markov model), students' future learning state is predicted, and dynamic adjustments are made according to students' learning state at different stages. For example, when students enter an "inefficient learning state", the system can push relevant learning resources in real time to help students resume their learning progress and adjust their learning state.

[0050] 3. Behavioral feature extraction and analysis In this embodiment, after completing the multimodal data fusion, the student behavior characteristics are further extracted and multi-dimensional analysis is performed: Online behavior feature extraction: Based on the students’ behaviors on the online learning platform (such as study time, homework submission, test scores, etc.), relevant behavioral features are extracted through feature engineering. Including but not limited to average study time, homework completion, online interaction frequency, test score distribution, etc. Each feature reflects the students’ online learning habits and learning effects.

[0051] Offline behavior feature extraction: Corresponding feature extraction is performed on students' offline classroom behavior data (such as classroom interaction frequency, experimental operation performance, etc.), mainly including the number of classroom interactions, experimental operation accuracy, and completion of classroom assignments.

[0052] Emotional fluctuation feature extraction: The emotion analysis model is used to extract the characteristics of students’ emotional fluctuations during the learning process, mainly including the positive and negative tendencies of emotions, the frequency of emotional fluctuations, and emotional stability. These characteristics are helpful in analyzing students’ learning motivation and psychological state.

[0053] Environmental factor feature extraction: Environmental data mainly includes students’ device usage, network status, external noise, etc. These factors affect students’ learning efficiency. Feature extraction mainly includes network latency, bandwidth, device usage frequency, etc.

[0054] 4. Analysis of data fusion results In this embodiment, after completing data fusion and behavioral feature extraction, an integrated learning model (such as random forest, support vector machine) is used to train and analyze the fused data, extract key features, and combine the learning state model to evaluate the student's learning state. Based on the data fusion results, a more comprehensive analysis of each student's learning behavior can be made, thereby providing strong support for subsequent learning effect prediction. Step 3: Learn to build a portrait The main purpose of step S3 is to use the student learning status and behavior characteristics data obtained in step S2 to establish a learning effect prediction model. By analyzing the student's learning history and behavior characteristics and combining advanced machine learning algorithms, this step can predict the student's future learning effect, grades and possible learning obstacles. The learning effect prediction model is the basis for the formulation of personalized learning plans and provides a scientific basis for subsequent teaching interventions and personalized recommendations.

[0055] 1. Choose a predictive model framework In this embodiment, a variety of machine learning models are used for training and comparison for learning effect prediction, including regression models (such as linear regression, support vector regression SVR), ensemble learning models (such as random forest, gradient boosting tree GBDT), deep learning models (such as multi-layer perceptron MLP), etc. By comparing and tuning a variety of algorithms, the most suitable prediction model for this project is selected.

[0056] Regression Model: For predicting continuous data such as student grades and learning progress, regression algorithms are used to build models. The regression model takes students’ learning characteristics (such as learning time, interaction frequency, etc.) as input to predict students’ future grades or learning outcomes.

[0057] Ensemble learning model: Ensemble learning methods such as Random Forest and Gradient Boosted Tree (GBDT) are used to combine the prediction results of multiple decision trees to improve the accuracy and stability of the model. These models can handle high-dimensional data and effectively reduce overfitting problems.

[0058] Deep Learning Models: When the amount of data is large, deep learning models (such as multi-layer perceptron MLP) are used for training. Deep neural networks can provide more accurate predictions in scenarios with complex nonlinear relationships. Multi-layer perceptrons learn the complex features of data through multiple hidden layers, thereby improving prediction capabilities.

[0059] 2. Feature Engineering and Optimization In this embodiment, feature engineering is a key step to improve the prediction accuracy of the model. By optimizing the learning behavior characteristics of students, the performance of the model is enhanced. This includes but is not limited to: Data standardization and normalization: In order to ensure the dimensional consistency of different features during model training, each feature is standardized so that the data is within a unified range. This helps to improve the model convergence speed and avoid certain features from having too much influence on the model prediction results.

[0060] Time series feature processing: For time series data such as students’ learning progress, time series analysis methods are used to extract features, including the changing trend of historical grades, fluctuations in online learning time, etc. These features help the model to more accurately predict students’ future academic performance.

[0061] Weighted integration of emotional fluctuations and environmental factors: In view of the characteristics of emotional fluctuations and environmental factors, a weighting mechanism is designed in this embodiment to combine these characteristics with the characteristics of student learning behavior to comprehensively evaluate the impact of students' psychological state and learning environment on learning outcomes. In this way, these factors can be effectively incorporated into the model for prediction.

[0062] 3. Model training and evaluation After data preprocessing and feature engineering are completed, in this embodiment, the training data is input into the selected machine learning model, and the model is evaluated through cross-validation. Cross-validation can ensure the generalization ability of the model and avoid instability of prediction results due to overfitting.

[0063] Training process: The model is trained multiple times using the training data set to optimize the model parameters until the model's prediction accuracy is optimal. For regression problems, the mean square error (MSE) is used as the loss function; for classification problems, the accuracy or F1 score is used as the evaluation criterion.

[0064] Model parameter adjustment: Through methods such as Grid Search or Random Search, the model's hyperparameters (such as tree depth, learning rate, etc.) are adjusted to further improve the model's prediction effect.

[0065] 4. Output and analysis of prediction results In this embodiment, after the model training is completed, the prediction results are analyzed to output the students' future learning effects, including the possible distribution of students' grades, possible learning obstacles, and predictions of learning efficiency. The prediction results can be displayed through visual charts, such as heat maps, line charts, etc., to help teachers quickly understand students' learning trends and possible weak links.

[0066] Through the above steps, step S3 successfully constructs an accurate learning effect prediction model, which combines students' behavioral data and learning status to provide effective support for personalized teaching intervention. This model can not only predict students' academic performance, but also identify learning obstacles in advance, thereby providing data support for subsequent teaching strategies.

[0067] Step 4: Prediction and intervention of learning effects The goal of step S4 is to make personalized teaching recommendations based on the learning situation of students and design appropriate teaching intervention plans based on the learning effect prediction model established in step S3. By combining factors such as students' learning status, emotional fluctuations, and external environment, the system will automatically generate personalized learning paths to help students overcome learning bottlenecks and improve learning effects. The core of this step is the accuracy and real-time nature of teaching interventions, ensuring that each student receives the right support at the right time.

[0068] 1. Personalized learning path generation In this embodiment, according to the student's learning status and prediction results, the system automatically generates a personalized learning path, which recommends appropriate learning resources and learning methods based on factors such as the student's learning progress, subject weaknesses, and emotional fluctuations.

[0069] Recommended learning content: Based on the student's performance in a specific subject, the system will recommend relevant learning materials. For example, for students with lower grades, the system will recommend basic materials and review questions; for students with faster progress in the subject, it will recommend advanced learning content to help them deeply understand the relevant knowledge points.

[0070] Recommended learning method: Based on students’ learning behavior data, the system will recommend strategies that best suit their learning style. For example, for students who like to learn through videos, the system will recommend relevant teaching videos; for students who like interactive learning, the system will push discussion-based online learning activities.

[0071] 2. Teaching intervention strategies In this embodiment, by real-time monitoring of students' learning situation, the system can automatically analyze whether students are in the learning bottleneck stage and provide intervention at the appropriate time. Intervention methods include but are not limited to: Real-time feedback and reminders: When the system detects that a student's learning has stagnated or his or her grades have dropped, it will immediately remind the student via email, text message or online message push, and provide the student with appropriate help resources.

[0072] Emotional intervention: For students with greater emotional fluctuations, the system detects the students' emotional state through the emotion analysis module and provides corresponding psychological counseling suggestions, such as relaxation exercises and positive emotion guidance.

[0073] Personalized Learning Support: When students are in an inefficient learning state, the system will automatically recommend group discussions, extracurricular tutoring and other support based on their learning history and emotional fluctuations to help students recover their learning state in time.

[0074] 3. Dynamically adjust learning paths and intervention plans As students' learning progress changes, personalized learning paths and intervention plans also need to be adjusted dynamically. The system monitors students' learning status and predicts results in real time, and adjusts learning resources and intervention strategies in a timely manner. For example, if students' learning efficiency is improved, the learning difficulty can be increased and more challenging learning content can be recommended; if students encounter learning bottlenecks, the learning path can be adjusted to reduce learning pressure.

[0075] Step 5: Personalized recommendations and real-time feedback The core goal of step S5 is to evaluate the effectiveness of the overall teaching strategy through comprehensive analysis of students' learning outcomes and teachers' teaching behaviors, and to provide data support for future teaching adjustments. This step forms a closed loop of teaching feedback and continuously optimizes the teaching process by providing feedback on students' learning outcomes and monitoring teachers' teaching activities. Data analysis and feedback mechanisms can help teachers identify problems in teaching in a timely manner and make adjustments to maximize teaching effectiveness.

[0076] 1. Student learning effect evaluation In step S5, a comprehensive evaluation of the students’ learning outcomes is first conducted, including but not limited to: Academic performance evaluation: Compare students' expected academic performance with their actual performance to assess their learning progress. By comparing the output of the prediction model with students' actual performance data, we can analyze students' achievement of learning goals and identify outstanding students and students with greater learning difficulties.

[0077] Behavioral analysis: By analyzing students' online learning behavior data (such as learning time, discussion frequency, video viewing progress, etc.), we can evaluate students' learning attitude and participation. Behavioral data can reflect students' learning input, help teachers understand whether students are truly engaged in learning, and adjust teaching strategies based on their behavior patterns.

[0078] Correlation analysis between emotional state and learning effect: In this embodiment, the correlation analysis between students' emotional fluctuations and learning effects is also performed. Through the emotional analysis module, students' emotional data is obtained in real time, and the relationship between it and learning effects is analyzed. If the emotional fluctuations are too large, it may affect students' academic performance. Teachers can make emotional interventions based on this information to improve students' learning effects.

[0079] 2. Teaching process and teaching quality evaluation While evaluating students’ learning outcomes, it is also necessary to monitor and evaluate teachers’ teaching process and teaching quality in real time, including: Teaching behavior monitoring: Teachers' teaching behaviors include classroom teaching, after-school tutoring, and the push of learning resources. This step collects and analyzes teachers' teaching behaviors through interactive data on the learning platform to evaluate their impact on students' learning outcomes. For example, if teachers interact with students less frequently, it may lead to a decrease in students' learning motivation. The system will remind teachers to strengthen interaction and enhance students' sense of participation.

[0080] Usage of teaching resources: This embodiment also monitors the use of teaching resources provided by teachers, including the frequency of use of courseware, teaching materials, videos, etc. and students' learning feedback. Based on these data, it evaluates whether the provided resources meet students' needs, whether the resources are outdated or not suitable for the current learning progress, and helps teachers adjust resource use strategies in a timely manner.

[0081] 3. Feedback on teaching strategies and suggestions for adjustment In this embodiment, the learning effect evaluation results of students and the teaching behavior analysis of teachers are integrated to provide teachers with systematic teaching feedback and adjustment suggestions. Including but not limited to: Improved teaching strategies: The system analyzes students’ learning outcomes and proposes targeted suggestions for adjusting teaching strategies. For example, if the analysis finds that some students have difficulties with specific content, the system will suggest that teachers provide targeted explanations for that part of the content or provide more auxiliary resources. If most students encounter learning bottlenecks in a certain link, the system will suggest that teachers increase the teaching time of that link or change the teaching method in the next teaching session.

[0082] Emotional and behavioral interventions: If the system detects that most students have large emotional fluctuations and that emotional fluctuations are related to academic performance, the system will suggest that teachers adopt appropriate emotional regulation methods, such as increasing classroom interaction and conducting group cooperative learning, to help students restore their positive learning emotions. Teachers can also adjust the classroom atmosphere and teaching rhythm to adapt to students' emotional needs and optimize learning outcomes.

[0083] Personalized coaching advice: For students with poor learning results or slow progress, the system will provide personalized tutoring suggestions, and teachers can adjust the content and methods of tutoring based on the system's suggestions. For example, if a student performs poorly in a certain learning link, the system will recommend tutoring plans, such as setting up after-school tutoring, conducting group discussions, etc., to help students overcome learning difficulties.

[0084] 4. Continuous monitoring and optimization Step S5 of this embodiment is not completed in one go, but is a continuous optimization process. The system will regularly monitor and provide feedback on the teaching effect, and update the teaching strategy based on new data. The specific process is as follows: Periodic data updates: The system continuously optimizes the teaching process by periodically updating student learning data, teaching behavior data, and sentiment analysis data. Each round of data update will be adjusted based on the feedback from the previous round to ensure the effectiveness of the teaching strategy.

[0085] Dynamically optimize teaching content: The system will dynamically adjust the teaching content based on students’ learning progress and teachers’ feedback. For example, if the teaching effect of a certain learning unit is not ideal, the system will automatically adjust the difficulty of the content, teaching methods, etc. to meet students’ learning needs.

[0086] Feedback mechanism between teachers and students: In addition to the system automatically providing feedback, it also allows teachers and students to interact and provide feedback. Teachers can make adjustments to the system based on their own experience, and students can provide feedback based on their own learning experience. The system will continuously optimize the teaching plan based on this feedback.

[0087] Through the above steps, step S5 successfully achieves comprehensive evaluation and continuous optimization of teaching effects. The system proposes accurate teaching adjustment suggestions by comprehensively considering students' learning performance, teachers' teaching behaviors and emotional fluctuations, thereby effectively improving the overall teaching quality and students' learning effects. Through this continuous feedback and optimization mechanism, it ensures that teaching strategies can be continuously adjusted according to students' needs and learning situations, and realizes the long-term effectiveness of personalized teaching.

[0088] Embodiment 1: Scene background: Suppose a clinical teaching course at a dental school adopts a hybrid online and offline teaching model. The course involves basic knowledge learning (such as dental anatomy, oral disease theory, etc.) and practical operation courses (such as tooth restoration technology, oral surgery, etc.). Students learn theoretical knowledge through the online learning platform and conduct practical operations and discussions in class. In this process, the system not only tracks the students' learning situation, but also adjusts the teaching strategy in real time based on the students' behavior and emotional feedback.

[0089] Based on the teaching scenario, this embodiment simulates the data collection, analysis and feedback process in a teaching cycle, and optimizes the teaching strategy according to the feedback.

[0090] 1. Step S1: Data collection In the second embodiment, various data of students are first collected on the teaching platform, including but not limited to: Learning behavior data: students’ access records to online learning platforms (learning time, number of videos watched, number of courseware downloads, homework submissions, etc.).

[0091] Offline classroom data: student participation in class (such as the number of questions asked, interaction frequency, experimental operation records, etc.), obtained through the classroom monitoring system and the teacher's teaching behavior records.

[0092] Emotional data: Collect students’ emotional feedback in real time through classroom interactive tools (such as electronic whiteboards and online discussion platforms) to understand students’ emotional reactions to course content.

[0093] 2. Step S2: Data preprocessing After data collection, all raw data enter the data preprocessing stage. The specific steps include: Cleaning and screening: Remove irrelevant data (such as invalid online learning records, repeated emotional feedback data, etc.) and screen out data related to student learning outcomes.

[0094] Data normalization and standardization: Normalize different types of data (such as learning time, sentiment scores, etc.) to ensure the comparability of various data sources.

[0095] Missing data filling: Use machine learning algorithms (such as interpolation, KNN, etc.) to fill in missing values ​​in student behavior data.

[0096] 3. Step S3: Feature extraction and modeling Based on the preprocessed data, feature extraction and model building are performed. In this embodiment, the system extracts the following important features based on the student's learning progress, classroom performance and emotional state: Learning progress characteristics: including students’ learning progress, homework completion, test scores, etc. on the online learning platform.

[0097] Classroom interaction characteristics: Analyze students’ interactions in class through the classroom feedback system, such as the number of questions asked, the frequency of interaction with teachers, the completion of experimental operations, etc.

[0098] Emotional fluctuation characteristics: Based on students’ emotional feedback in the course (such as positive emotions, negative emotions, frequency of emotional fluctuations, etc.), extract students’ emotional fluctuation characteristics.

[0099] Based on these features, the system uses machine learning algorithms (such as decision trees, random forests, etc.) to build a learning effect prediction model. The model will predict students' learning outcomes and emotional fluctuations throughout the course cycle, helping teachers understand students' learning status.

[0100] 4. Step S4: Learning effect prediction and analysis After completing the model training, the system starts to predict the learning effect in real time. Based on the real-time data input of students, the model can predict the students' academic performance, operational ability and emotional change trends. Specifically, it includes: Academic performance prediction: Predict students’ final exam scores and course evaluation based on their online learning scores, homework submissions, and classroom performance.

[0101] Prediction of operational ability: For oral skills courses, students’ surgical skill level is predicted based on their performance in experimental operations.

[0102] Emotional state prediction: By analyzing the relationship between students' emotional fluctuations and academic performance, the potential impact of students' emotional state on their academic performance is predicted.

[0103] 5. Step S5: Teaching feedback and strategy adjustment Based on the prediction of learning effects, the system provides teachers with teaching feedback and strategy adjustment suggestions. Specifically, it includes: For students with poor academic performance: The system finds that some students' online learning performance is far from their actual performance, and they are not active in class. The system recommends that teachers increase extracurricular tutoring time for these students or recommend additional learning resources to make up for their learning weaknesses.

[0104] For students with insufficient operational ability: The system detected that some students in the experimental course had insufficient operational ability, especially poor performance in complex operational links. The system suggested that teachers arrange individual guidance or group practice for these students, and help students overcome difficulties through one-on-one tutoring.

[0105] For students with large emotional fluctuations: The system also identifies that some students have large emotional fluctuations, especially those who frequently experience emotional fluctuations (such as anxiety and depression) in class. The system prompts teachers to conduct emotional counseling in a timely manner, for example, through more encouraging words or moderately slowing down the teaching pace, to ease students' emotions and improve their learning effects.

[0106] Teachers’ teaching behavior feedback: The system also provides feedback to teachers based on students’ data, pointing out the negative impact of certain teaching behaviors (such as too monotonous teaching content or lack of interaction) on students’ learning outcomes. The system recommends that teachers add interactive teaching sessions in subsequent classes, such as group discussions and case analysis, to improve students’ learning participation.

[0107] Embodiment 2: Scene background: A medical university has opened a basic internal medicine course, which covers the basic theories and clinical knowledge of various human systems. The course adopts a hybrid online and offline teaching mode. Online learning includes theoretical courses and some auxiliary materials, while offline classes focus on explaining difficult points, conducting practical operations and case analysis. In this course, the system not only collects students' learning behavior data, but also analyzes students' emotional state, so as to adjust teaching strategies and optimize teaching effects in real-time feedback.

[0108] Based on this teaching scenario, this embodiment predicts students' learning outcomes through learning data analysis, helps teachers improve their teaching methods, and makes targeted teaching adjustments in a timely manner.

[0109] 1. Step S1: Data collection In this embodiment, data collection is first performed, mainly including the following types of data: Online learning data: Students’ course access records on the learning platform, including study time, video viewing times, post-class test scores, etc.

[0110] Students submit assignments, papers or reports on the platform and record the completion quality of each task.

[0111] Participation in forums or discussion boards, such as posting, commenting, liking, and other interactive behaviors.

[0112] Offline classroom data: The level of student participation in class, including raising hands to answer questions, classroom interaction, and experimental operations.

[0113] Students' class attendance records are used to detect whether students actively participate in class discussions and practices.

[0114] Data collection of classroom interaction tools to monitor the frequency and quality of real-time interaction between students and teachers.

[0115] Sentiment data: Students’ emotional states are collected in real time through classroom interactive tools, such as online questionnaires, emotional surveys, and real-time emotion recognition tools.

[0116] After each class, students can provide emotional feedback on the class content, marking their liking for the course, confusion, anxiety, etc.

[0117] 2. Step S2: Data preprocessing After data collection, data preprocessing is performed, including the following steps: Data cleaning: Remove invalid data (such as data from erroneous operations or system anomalies) to ensure that only meaningful data is retained.

[0118] Data uniform formatting: Format data from different sources for uniform processing and analysis. For example, align the timestamps of emotional feedback data and learning behavior data to ensure data consistency.

[0119] Data missing processing: For missing data, data interpolation technology (such as KNN algorithm, interpolation method, etc.) is used to fill the missing values ​​to ensure data integrity.

[0120] 3. Step S3: Feature extraction and modeling In this embodiment, the goal of feature extraction is to better understand the behavior patterns of students in class and the impact of their emotional states on learning outcomes. The main steps include: Learning behavior feature extraction: Study time: Calculate the study time invested by each student on the platform as a characteristic of learning enthusiasm and initiative.

[0121] Learning progress: Based on the students’ online learning progress, analyze how much course content they have completed and compare it with the total course content to derive a learning progress indicator.

[0122] Test scores: Analyze students’ performance in periodic tests, perform weighted average calculations, and extract each student’s test scores.

[0123] Classroom interaction feature extraction: Student class participation: The degree of student activity in class can be determined through data such as classroom questions and answers, discussion interactions, and experimental operations.

[0124] Student classroom behavior: Extract students’ learning behavior characteristics in class through classroom behavior data recorded by teachers (such as the number of times they speak, the number of questions asked, etc.).

[0125] Emotional feature extraction: Emotional fluctuation analysis: Through students’ emotional feedback in class, the emotional fluctuation index is extracted to evaluate students’ emotional stability throughout the semester.

[0126] Emotional and academic performance association: Through statistical analysis, the correlation between students' emotional states and their academic performance was evaluated.

[0127] Based on these feature data, machine learning algorithms (such as support vector machine SVM, random forest, etc.) are used to build a learning effect prediction model to predict students' final grades, learning attitudes, and operational capabilities.

[0128] 4. Step S4: Learning effect prediction and analysis Use the constructed learning effect prediction model to predict each student's academic performance, classroom performance, and emotional state. The system will automatically calculate the following: Academic performance prediction: Based on students’ learning progress, homework grades, test performance and other data, the system predicts students’ scores in the final exam and generates a distribution chart of academic performance.

[0129] Operational ability prediction: For practical operation courses, the system predicts students' actual operational ability, such as the ability to handle actual cases, by analyzing their performance in simulated experiments.

[0130] Emotional prediction: The system will also predict students' emotional fluctuations, especially those with larger emotional fluctuations. The system will give warnings and prompt teachers to take timely intervention measures.

[0131] These prediction results will help teachers better understand students’ learning situation and make timely teaching adjustments.

[0132] 5. Step S5: Teaching feedback and strategy adjustment According to the prediction results of step S4, the system provides teachers with targeted teaching feedback and strategy adjustment suggestions, including: For students with low academic performance: The system identifies that some students have low predicted academic performance and gives targeted teaching suggestions, such as arranging additional online tutoring and giving individual guidance.

[0133] For students with large emotional fluctuations: The system finds that some students have large emotional fluctuations, and emotionally unstable students may affect learning outcomes. The system recommends that teachers strengthen emotional counseling, such as by increasing interactive sessions, adjusting the pace of teaching, and other methods to improve students' emotional states.

[0134] For students with poor practical ability: The system detects that some students have weak experimental operation ability. The system recommends that teachers arrange more practical operation opportunities for these students and adjust the difficulty of classroom content as needed.

[0135] In addition, the system also provides feedback on teachers’ teaching behavior. For example, if students’ overall scores are low, the system may suggest that teachers increase the explanation of difficult points or adjust the interactive teaching method to ensure that students can better understand complex knowledge points.

[0136] By implementing this solution, teachers can grasp students' learning progress, emotional fluctuations and classroom performance in real time, and adjust teaching strategies based on these data. With the help of learning situation prediction, teachers can identify students with learning difficulties and provide them with personalized tutoring, thereby improving students' academic performance and operational ability.

[0137] This embodiment not only improves the precision of classroom management, but also enhances the emotional interactivity of the teaching process through accurate emotional analysis. Ultimately, the students' learning experience in the classroom is optimized and the teaching effect of the course is effectively improved.

[0138] Embodiment three: Scene background: A vocational and technical college has opened a computer programming course to cultivate students' programming ability and practical skills. The course adopts a hybrid online and offline teaching mode, combining online learning platforms, classroom teaching, laboratory practice and other methods to comprehensively improve students' programming ability and ability to solve practical problems. Teachers hope to optimize teaching strategies and adjust teaching rhythm in time through learning situation analysis and learning effect prediction, so as to improve students' learning effect and practical ability.

[0139] In this embodiment, the system not only helps teachers analyze students' learning behaviors, but also helps teachers provide personalized tutoring to students and improve teaching content and methods by predicting students' programming abilities and learning outcomes.

[0140] 1. Step S1: Data collection In this embodiment, data collection covers students' online learning data, offline classroom data, programming practice data, and emotional feedback data, specifically including the following aspects: Online learning data: The frequency of students logging into the platform, online learning time, video viewing progress and chapter completion.

[0141] The quality of each programming task submission, the time and quality score of the homework completion.

[0142] Scores on end-of-course quizzes, online tests, and completion of self-study modules.

[0143] Online programming experiments and code submission records monitor students' actual programming ability performance on the online programming platform.

[0144] Offline classroom data: Students participate in programming discussions, case analysis, and real-world problem solving in class.

[0145] Student contribution to group projects, teamwork, and quality of completed projects.

[0146] Records of students’ classroom interactions, such as the number of questions asked, the quality of answers, and their enthusiasm for participating in discussions.

[0147] Programming practice data: Students' specific operations in programming practice, such as the quality of written code, compilation error rate, and debugging problem-solving ability.

[0148] Project completion status, programming efficiency, function realization status, etc. in the laboratory.

[0149] Sentiment data: Students’ emotional feedback on online learning modules and classroom teaching content.

[0150] Students' emotional fluctuations, such as anxiety, satisfaction, and interest in programming tasks.

[0151] 2. Step S2: Data preprocessing The data preprocessing steps in this embodiment include the following: Data cleaning: Remove invalid or erroneous data, such as duplicate records, blank data, or invalid operations, to ensure data accuracy.

[0152] Data standardization: Standardize learning behavior data from different sources. For example, standardize online learning time, class participation times, programming task completion, and other indicators to make the data comparable.

[0153] Missing data filling: For some missing data, interpolation, mean filling and other methods are used to fill them to ensure the smooth progress of subsequent analysis.

[0154] 3. Step S3: Feature extraction and modeling In this embodiment, the focus of feature extraction is to build a prediction model through students' behavioral data, programming ability and emotional state. The main steps are as follows: Learning behavior feature extraction: Learning time: Extract students’ learning engagement characteristics based on their online learning time and classroom learning time.

[0155] Programming task completion: Analyze students’ programming ability and learning progress based on their scores, submission quality, and task completion time.

[0156] Participation: Combine students’ classroom interaction and teamwork to extract students’ participation characteristics.

[0157] Programming ability feature extraction: Code quality: Extract students’ programming ability characteristics based on the quality of the code they submit (such as whether it compiles, code complexity, error rate, etc.).

[0158] Debugging ability: Analyze students’ performance during debugging to determine whether they can quickly identify and fix errors.

[0159] Project implementation: Students’ actual programming ability will be assessed based on their performance in practical projects.

[0160] Emotional feature extraction: Emotional Fluctuation Index: By analyzing students’ emotional feedback on course content, the emotional fluctuation index is extracted to evaluate students’ emotional state.

[0161] Emotional stability: Analyze students’ emotional changes throughout the learning process, evaluate their emotional stability, and then predict their learning outcomes.

[0162] 4. Step S4: Learning effect prediction and analysis Based on the extracted feature data, a learning effect prediction model is constructed using machine learning algorithms (such as support vector machine SVM, XGBoost, neural network, etc.). The specific prediction contents include: Academic performance prediction: Predict students’ final academic performance based on factors such as their learning behavior data, programming task performance, and emotional fluctuations.

[0163] Programming ability prediction: Predict students’ programming ability based on their performance in programming tasks and their actual operational ability in experimental projects.

[0164] Emotional stability analysis: Through students’ emotional data, predict the emotional fluctuations that may occur in students during the learning process and give corresponding intervention suggestions.

[0165] The prediction results will be presented in the form of intuitive charts to help teachers effectively evaluate students' learning status.

[0166] 5. Step S5: Teaching feedback and strategy adjustment The system generates teaching feedback and strategy adjustment suggestions based on the learning effect prediction results, including: For students who are lagging behind: The system will flag students who are lagging behind and recommend teachers to provide them with personalized tutoring, extend their study time or arrange additional online tutoring.

[0167] For students with weak programming skills: If the system identifies that some students have low scores in programming tasks or weak debugging skills, the system will recommend that teachers arrange targeted programming exercises and improve the practical aspects of project practice.

[0168] For students with greater emotional fluctuations: For students with greater emotional fluctuations, the system will suggest that teachers reduce students' anxiety and improve their participation in learning by increasing classroom interaction, adjusting the teaching pace, etc.

[0169] 6. Step S6: Personalized teaching adjustment Based on students’ learning progress, programming ability, emotional state and other data, the system provides teachers with personalized teaching adjustment suggestions, including: Course content adjustment: Based on the predicted learning outcomes of students, the system may suggest that teachers adjust the depth or difficulty of course content to ensure that the course content matches the students’ learning progress.

[0170] Adjustment of teaching methods: After systematically analyzing students’ emotional feedback data, teachers may be advised to increase interactive sessions, optimize teaching strategies, and improve students’ learning interest and emotional stability.

[0171] Adjustment of tutoring strategies: In view of students’ weaknesses in programming skills, the system may suggest that teachers provide more personalized tutoring, such as arranging extracurricular programming exercises and guiding students in project practice.

[0172] By implementing this solution, teachers can grasp students' learning status, programming ability and emotional fluctuations in real time, and adjust teaching strategies in time according to these data. Students' learning effect in the course has been significantly improved, and their programming ability and practical ability have been strengthened. The introduction of emotional fluctuation analysis also helps teachers pay attention to students' emotional needs and ensure the balance and stability of the teaching process.

[0173] This embodiment not only improves the level of personalization of teaching, but also improves the accuracy of teaching strategies through data-driven decision support, so that students can achieve optimal learning effects under personalized teaching support.

[0174] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching, characterized in that: The following steps are involved: Step 1: Collect students’ online learning data, offline classroom data, emotional data, and environmental data; Step 2: Preprocess the collected multi-source data, remove outliers and perform standardization; Step 3: Fuse multimodal data to extract characteristics of students’ learning behaviors, emotional fluctuations, and academic performance; Step 4: Construct a student's learning profile, including academic ability profile, learning behavior profile, emotional profile, and learning resource preference profile; Step 5: Based on the learning profile and student historical data, predict learning outcomes and provide real-time feedback and intervention strategies.

2. According to claim 1, the method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching is characterized in that: The pre-processing step comprises: De-noise the learning behavior data to remove invalid information; Use sliding window method to process time series data; The sentiment dictionary analysis method is used to extract the emotional fluctuation characteristics of the emotional data; Academic performance data were standardized to eliminate dimension effects.

3. The method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching according to claim 1 is characterized in that: The multimodal data fusion step includes: Jointly model learning behavior, emotional data, and academic performance data using cross-attention mechanisms; Different deep learning models are used for each data modality, including LSTM for learning behavioral data, BiLSTM for sentiment data, and XGBoost for academic performance data.

4. The method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching according to claim 1 is characterized in that: The steps of constructing the learning portrait include: Build academic ability portraits through knowledge tracking models and update students’ knowledge points in real time; Use clustering algorithms to classify learning behaviors, identify students’ learning styles, and generate learning behavior portraits based on the styles; Analyze students’ emotional fluctuations based on emotional data and generate emotional portraits; Generate a learning resource preference profile based on learning history and identify the types of learning resources preferred by students.

5. The method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching according to claim 1 is characterized in that: The learning effect prediction step includes: Use the Transformer model to predict students’ learning progress and capture long-term dependencies; Based on the multi-task learning model, multiple tasks such as academic performance, emotional fluctuations, and learning attitude are predicted simultaneously; Weighted voting and Stacking models are used to fuse the prediction results of different data sources.

6. The method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching according to claim 1 is characterized in that: The real-time feedback step includes: Adjust the difficulty of learning tasks in real time according to students’ learning progress and emotional fluctuations; Provide personalized learning resources and learning task recommendations for students who are lagging behind in their studies; Generate encouraging feedback based on students' emotional states.

7. The method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching according to claim 1, characterized in that: Intervention strategies described include: Providing targeted remedial materials and exercises to students with poor academic performance; Recommend emotion management techniques and psychological counseling resources to students with greater emotional fluctuations; Students with passive learning styles are motivated to learn through motivational content and interactive design.

8. The method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching according to claim 1, characterized in that: The personalized learning path recommendation step includes: Dynamically recommend appropriate learning paths based on students’ learning profiles and learning styles; According to changes in learning behavior data and emotional portraits, the learning path is adjusted in real time, and customized learning resources are pushed.

9. The method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching according to claim 1, characterized in that: The learning resource recommendation step includes: Based on students’ learning history data, collaborative filtering algorithms are used to recommend learning resources for students with similar learning styles. Dynamically adjust resource recommendations based on students’ emotional portraits.

10. The method for analyzing learning situation and predicting learning effect of online and offline hybrid teaching according to claim 1, characterized in that: The real-time feedback and intervention steps are implemented through a learning management system, which includes a learning data acquisition module, a data preprocessing module, a learning effect prediction module, a real-time feedback module and a personalized recommendation module. The modules interact in real time through the network to form a closed-loop feedback mechanism.

Images