Intelligent screening method and system for student psychological state recognition

Abstract

The invention discloses an intelligent screening method and system for student psychological state recognition, and belongs to the technical field of data analysis. Student psychological log data in a continuous time period is obtained and subjected to standardized cleaning, an emotion fluctuation amplitude feature and a context relevance feature are extracted, and an emotion coherence index is comprehensively calculated based on the emotion fluctuation amplitude feature and the context relevance feature; the complexity of emotion evolution is quantified; furthermore, the classification accuracy of a traditional model is evaluated in different emotion coherence index intervals, if it is detected that the accuracy is lower than a preset threshold value in a high-complexity interval, the data input quantity reflecting time sequence features is dynamically adjusted, and iterative training and optimization are conducted on the adjusted model. The detection accuracy and intervention timeliness of a psychological state recognition system in a complex emotion change scene can be remarkably improved, false negative is effectively reduced, and the intelligence and precision level of psychological health risk screening is improved.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a method and system for intelligently screening students' psychological states. Background Art

[0002] Intelligent screening for student mental state identification refers to the use of artificial intelligence technologies (such as big data analysis, machine learning, natural language processing, etc.) to automatically detect and analyze students' mental health status through questionnaires, voice, text, behavioral data, and other methods, to promptly identify potential risks of psychological problems, and help schools and professionals conduct early intervention and scientific management, thereby more effectively promoting students' mental health development.

[0003] The existing technology has the following shortcomings:

[0004] In a long-term student psychological status tracking system, using traditional machine learning models (such as support vector machines (SVMs)) to process a week's worth of psychological diary text is inadequate to model complex temporal and contextual dependencies. Consequently, the system can only make static decisions based on short, localized text, ignoring the coherent trends of emotional change. As a result, while the system appears to have high classification accuracy, a large number of potential psychological crises are not identified in a timely manner, resulting in significant false negatives, reducing the effectiveness of interventions and easily leading to the serious consequence of missing long-term emotional deterioration. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for intelligent screening of students' psychological state recognition to address the deficiencies in the background technology.

[0006] In order to achieve the above-mentioned object, the present invention provides the following technical solution: a method for intelligently screening students' psychological states, comprising:

[0007] Acquiring student psychological diary data within a continuous time period, and cleaning the psychological diary data to form a standardized text data set;

[0008] Extracting emotion fluctuation amplitude features and context relevance features from the text dataset;

[0009] Calculating an emotional coherence index based on the emotional fluctuation amplitude feature and the context relevance feature, wherein the emotional coherence index is used to indicate the complexity of emotional evolution;

[0010] Based on the trained traditional machine learning model, the students' psychological states were classified and predicted, and the classification accuracy was statistically analyzed within different emotional coherence index ranges.

[0011] Determining whether the classification accuracy of the traditional machine learning model within the high emotion coherence index range is lower than a preset accuracy threshold; if so, determining that the model is insufficient in capturing the contextual features of the time-series psychological data;

[0012] When it is determined that the model accuracy is insufficient, the amount of data input that reflects the time series characteristics is dynamically adjusted, and the adjusted traditional machine learning model is iteratively trained and evaluated.

[0013] Preferably, the method for obtaining the emotion change amplitude value is: for each log text, apply the sentiment analysis model to generate the emotion score, which is recorded as , where t represents the time node, and the amplitude of a single emotion change is calculated based on the absolute value of the difference in emotion scores between two consecutive time nodes , the expression is: ; Calculate the average emotion change amplitude in a continuous time period. The calculation expression is: ; Where N is the total number of time nodes, and the calculation expression for the emotion change amplitude value ME is: .

[0014] Preferably, the method for obtaining text association similarity is: converting the log text of each time node into a vector representation of fixed dimension through the embedding model, which is recorded as , calculate the cosine similarity between the text vectors of two consecutive time nodes as the single text association similarity , the calculation expression is: ;in: is the modulus of each vector, and the text association similarity is calculated: ; M is the total number of time points for obtaining text vectors, is the text association similarity, which is used to indicate the degree of context coherence.

[0015] Preferably, an emotional coherence index is calculated based on the emotional fluctuation amplitude feature and the context relevance feature. The emotional coherence index is used to represent the complexity of emotional evolution, specifically including:

[0016] The emotion change amplitude value and text association similarity are normalized so that they are both between [0, 1]. The emotion coherence index is calculated based on the normalized emotion change amplitude value and text association similarity.

[0017] Preferably, the byte comparison deviation index and the check value comparison anomaly index are normalized so that they are both between [0, 1], and the normalized byte comparison deviation index and the check value comparison anomaly index are weighted averaged and summed to obtain the accuracy score of the current message parsing.

[0018] Preferably, the student's psychological state is classified and predicted based on the trained traditional machine learning model, and the classification accuracy is statistically analyzed in different emotional coherence index ranges, specifically including:

[0019] Using a dataset with labeled mental state categories, a traditional machine learning model is used for model training. Input features include emotion scores, emotional tendencies, and behavioral pattern indicators, and the corresponding mental state category labels are output.

[0020] Input the psychological diary data of the students to be tested into the trained model to automatically classify and predict their psychological states, and obtain the predicted labels for each student in each time period;

[0021] For each student, the emotional coherence index (ECI) was calculated based on their psychological diary for consecutive time periods;

[0022] All students were divided into different intervals according to their ECI values, including: low complexity interval: ECI < 0.3; medium complexity interval: 0.3 ≤ ECI < 0.6; high complexity interval: ECI ≥ 0.6; each interval contains several student samples;

[0023] In each ECI interval, the classification accuracy of the model is calculated separately:

[0024] Define whether the prediction of each sample is correct;

[0025] Count the number of samples predicted correctly in each interval ;

[0026] Count the total number of samples in each interval ;

[0027] Calculate interval accuracy , the expression is: .

[0028] Preferably, determining whether the classification accuracy of the traditional machine learning model in the high emotion coherence index range is lower than a preset accuracy threshold, and if so, determining that the model has insufficient ability to capture the contextual features of the time-series psychological data, specifically includes:

[0029] The classification accuracy threshold θ is preset. Among all student samples, the sample set with the emotional coherence index ECI ≥ 0.6 is screened out and recorded as the high ECI interval dataset Dhigh-ECI;

[0030] In Dhigh-ECI, the actual accuracy of traditional machine learning models in predicting mental state classification is statistically analyzed. , and its calculation formula is: ;in: Indicates the number of correctly classified samples in the high ECI interval, Indicates the total number of samples in the high ECI interval;

[0031] Compare the classification accuracy in the high ECI interval with the preset threshold θ: ≥θ, the model is considered to have good classification ability under complex emotional evolution; if <θ, it is determined that the model is insufficient in capturing the contextual features of temporal psychological data.

[0032] Preferably, when it is determined that the model accuracy is insufficient, the amount of data input reflecting the time series characteristics is dynamically adjusted, specifically including:

[0033] Apply DTW calculation to each student's continuous emotion score sequence or text sentiment tendency sequence, and dynamically adjust the input features according to the degree of change of DTW features:

[0034] If the DTW distance is greater than the preset distance threshold, the data input dimension is increased;

[0035] If the DTW distance is less than or equal to the preset distance threshold, the input amount is reduced;

[0036] Combine the new time series features adjusted based on DTW with the original static features to form a new training dataset;

[0037] Retrain traditional machine learning models using updated datasets.

[0038] The present invention also provides an intelligent screening system for identifying students' psychological states, comprising a data acquisition and preprocessing module, a feature extraction module, an emotion coherence calculation module, a psychological state identification and performance evaluation module, an accuracy detection and deficiency judgment module, and a dynamic adjustment and iterative optimization module;

[0039] Data collection and preprocessing module: obtains students' psychological diary data within a continuous time period, cleans the psychological diary data, and forms a standardized text data set;

[0040] Feature extraction module: extracting emotion fluctuation amplitude features and context relevance features from the text dataset;

[0041] Emotional coherence calculation module: calculates an emotional coherence index based on the emotional fluctuation amplitude feature and the context relevance feature, wherein the emotional coherence index is used to represent the complexity of emotional evolution;

[0042] Mental State Recognition and Performance Evaluation Module: This module classifies and predicts students' mental states based on the trained traditional machine learning model, and calculates the classification accuracy within different emotional coherence index ranges.

[0043] Accuracy detection and deficiency judgment module: determines whether the classification accuracy of the traditional machine learning model in the high emotional coherence index range is lower than a preset accuracy threshold. If so, it is determined that the model is insufficient in capturing the contextual features of the time-series psychological data;

[0044] Dynamic adjustment and iterative optimization module: When it is determined that the model accuracy is insufficient, the data input amount reflecting the time series characteristics is dynamically adjusted, and the adjusted traditional machine learning model is iteratively trained and evaluated.

[0045] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0046] 1. By incorporating emotional fluctuation amplitude and contextual relevance features, this paper innovatively proposes an emotional coherence index as a key metric for measuring the complexity of emotional evolution. Combined with classification accuracy assessment within the emotional coherence index range, this method dynamically detects performance bottlenecks in traditional machine learning models when processing complex time-series psychological data. When insufficient accuracy is detected in highly complex emotional evolution samples, this paper utilizes a dynamic time warping algorithm to intelligently adjust the input of time series features and iteratively optimize the traditional model. This effectively improves the sensitivity and accuracy of the psychological state recognition system to potential psychological crises, significantly reduces the false negative rate, and enhances the early warning capabilities and practical value of mental health screening.

[0047] 2. The intelligent student mental state identification and screening system proposed in this invention integrates a data acquisition and preprocessing module, a feature extraction module, an emotion coherence calculation module, a mental state identification and performance evaluation module, an accuracy detection and deficiency judgment module, and a dynamic adjustment and iterative optimization module. It can dynamically and adaptively adjust during data interaction and frame parsing, improving the accuracy and stability of data processing and ensuring the integrity and reliability of student mental state data transmission. Overall, this invention not only compensates for the existing system's insufficient ability to model the dynamic evolution of emotions, but also improves the intelligence level, dynamic adaptability, and security of the intelligent mental state screening system through system-level linkage optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0049] Figure 1 This is a mind map of the method of the present invention.

[0050] Figure 2This is a mind map of the system modules of the present invention. DETAILED DESCRIPTION

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of 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. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0052] Example 1, please refer to Figure 1 As shown, the intelligent screening method for identifying students' mental states described in this embodiment includes:

[0053] Acquiring student psychological diary data within a continuous time period, and cleaning the psychological diary data to form a standardized text data set;

[0054] Extracting emotion fluctuation amplitude features and context relevance features from the text dataset;

[0055] Calculating an emotional coherence index based on the emotional fluctuation amplitude feature and the context relevance feature, wherein the emotional coherence index is used to indicate the complexity of emotional evolution;

[0056] Based on the trained traditional machine learning model, the students' psychological states were classified and predicted, and the classification accuracy was statistically analyzed within different emotional coherence index ranges.

[0057] Determining whether the classification accuracy of the traditional machine learning model within the high emotion coherence index range is lower than a preset accuracy threshold; if so, determining that the model is insufficient in capturing the contextual features of the time-series psychological data;

[0058] When it is determined that the model accuracy is insufficient, the amount of data input that reflects the time series characteristics is dynamically adjusted, and the adjusted traditional machine learning model is iteratively trained and evaluated.

[0059] Set a fixed data collection period, such as daily, to continuously collect student psychological diary data for 7, 14, or 30 days. Include timestamp information for each data point to ensure that the diary has a clear chronological order.

[0060] Collect students' self-reported psychological diaries through designated channels, such as: online psychological diary system (webpage, mini program); mobile application psychological recording module; campus mental health management platform.

[0061] Standardize the content submitted by students, requiring them to write psychological feelings, daily emotions, stress experiences, descriptions of important events, etc. in natural language. Set a minimum word count requirement (e.g., a daily log should be no less than 100 words) to ensure sufficient data volume.

[0062] Uniformly store the log data in a standard text format (such as a txt or json document encoded in UTF-8). Bind the text content to the corresponding timestamp to generate structured data units;

[0063] Exclude the following invalid logs: empty text or insufficient content (such as the number of characters is below the threshold, e.g., less than 30 characters); obviously irrelevant content (such as only containing emojis, a large amount of garbled characters, advertising information); non-natural language content (such as pure numbers, meaningless strings of letters).

[0064] Word segmentation and basic processing (process for Chinese / English separately): Chinese: Use Jieba or BERT tokenizer to split sentences into meaningful words or phrases. English: Adopt standard natural language tools (such as NLTK, spaCy) for Tokenization. Remove stop words (such as "的", "了", "and", "the", etc. words without actual emotional meaning).

[0065] Calculate the basic statistical features of each log text (such as the number of words, sentiment score, number of sentences). Detect data samples with extremely abnormal features (such as the number of words far exceeding the normal range or extreme values of emotion scores), and conduct manual review or automatic exclusion.

[0066] Align the text logs by day, hour, or other set granularity, and fill in the missing time nodes (for example, fill the date without logs with the identifier "no data").

[0067] Form a unified data table or data packet format, anonymize the personal information of students (such as name, student ID), and use an anonymous ID for identification. Ensure that the log data complies with the requirements of data protection regulations. <​​​​​​​​​ ; Calculate the average emotion change amplitude in a continuous time period. The calculation expression is: Where N is the total number of time nodes, and the calculation expression for the emotional change amplitude value ME (which measures the discreteness of emotional fluctuations) is: .

[0070] A larger value for emotional change indicates more dramatic mood swings within a continuous timeframe, with significant emotional instability. This often indicates a complex psychological state, potentially experiencing a shift in multiple emotional experiences, such as anxiety, joy, and depression. This complex and unpredictable emotional evolution suggests the need for further attention to potential psychological risks.

[0071] Conversely, smaller emotional fluctuations indicate a more gradual change in the student's emotional score across time points, a more stable emotional state, and a simple, controllable evolution. In this case, the student's emotional state exhibits a continuous and consistent pattern, with low complexity in emotional evolution, whether consistently positive or negative. However, persistent negativity may indicate long-term mental health risks, requiring further analysis in conjunction with other characteristics.

[0072] The method for obtaining text association similarity is to convert the log text of each time node into a fixed-dimensional vector representation through an embedding model (such as BERT, Word2Vec, Sentence-BERT), which is recorded as Calculate the cosine similarity between the text vectors of two consecutive time nodes as the single text association similarity , the calculation expression is: ;in: is the modulus of each vector, and the text association similarity is calculated: ; M is the total number of time points for obtaining text vectors, is the text association similarity, which is used to indicate the degree of context coherence.

[0073] The greater the text-related similarity, the more semantically consistent the student's psychological diary content is between adjacent time points, the more coherent their emotional state and psychological expressions are, and the relatively smooth and orderly evolution process is. This generally indicates that the student's emotions fluctuate slightly, their psychological experiences are continuous, and they may be in a stable emotional state. Whether positive or negative, their emotional trajectory is relatively simple and predictable, and their evolution is less complex.

[0074] Conversely, lower text-related similarity indicates greater semantic differences between log texts at consecutive time points, with noticeable jumps in content, indicating frequent changes in emotional state or significant shifts in psychological expression style. In these cases, students may be experiencing a complex evolution of emotional discontinuities, sudden changes in psychological state, or a complex interweaving of multiple emotions. This results in weak emotional coherence and a high degree of emotional evolution, necessitating special attention to the risk of abnormal psychological fluctuations.

[0075] An emotional coherence index is calculated based on the emotional fluctuation amplitude feature and the context relevance feature. The emotional coherence index is used to indicate the complexity of emotional evolution, specifically including:

[0076] The emotion change amplitude value and text association similarity are normalized so that they are both between [0, 1]. The emotion coherence index is calculated based on the normalized emotion change amplitude value and text association similarity.

[0077] For example, the present invention can use the following formula to calculate the emotional coherence index, and the calculation expression is: Where, is the emotional coherence index, is the amplitude of emotion change, is the text association similarity, is the weight coefficient of the emotion change amplitude value and the text association similarity (which can be optimized based on experimental experience or machine learning), and Both are greater than 0.

[0078] Based on the trained traditional machine learning model, the students’ psychological states are classified and predicted, and the classification accuracy is statistically analyzed within different emotional coherence index ranges, including:

[0079] Using a dataset with labeled mental state categories (e.g., health, anxiety, depression), we train models using traditional machine learning models (e.g., support vector machines (SVMs), random forests (RFs), and XGBoost). Input features can include sentiment scores, emotional tendencies, behavioral pattern indicators, and the output is the corresponding mental state category label.

[0080] The psychological diary data of the students to be tested (after cleaning and feature extraction) is input into the trained model to automatically classify and predict their psychological states. The predicted labels for each student in each time period are obtained.

[0081] For each student, the emotional coherence index (ECI) was calculated based on their psychological diary for consecutive time periods according to the established method (combination of the emotional change amplitude value and text association similarity).

[0082] Divide all students into different intervals according to their ECI values, for example:

[0083] Low complexity range: ECI < 0.3;

[0084] Medium complexity range: 0.3≤ECI<0.6;

[0085] High complexity range: ECI ≥ 0.6;

[0086] Each interval contains several student samples.

[0087] In each ECI interval, the classification accuracy of the model is calculated separately. The specific method is as follows:

[0088] Define whether the prediction of each sample is correct;

[0089] Count the number of samples predicted correctly in each interval ;

[0090] Count the total number of samples in each interval ;

[0091] Calculate interval accuracy , the expression is: .

[0092] By comparing classification accuracy across different Emotional Coherence Index (ECI) ranges, we can assess the performance of traditional machine learning models in situations with varying degrees of emotional complexity. If classification accuracy drops significantly within high ECI ranges, we can further infer that the model's ability to capture complex emotional changes is insufficient.

[0093] Determine whether the classification accuracy of the traditional machine learning model within the high emotional coherence index range is lower than a preset accuracy threshold. If so, determine that the model is insufficient in capturing the contextual features of the time-series psychological data. Specifically, the following are included:

[0094] A preset classification accuracy threshold θ is set, such as 70%, 75%, or other standards based on business needs. The threshold is used to determine whether the model's performance meets the requirements under a specific level of emotional complexity.

[0095] Among all student samples, a sample set with an emotional coherence index (ECI) higher than the set limit (e.g., ECI ≥ 0.6) was selected and recorded as the high ECI interval dataset Dhigh-ECI.

[0096] In Dhigh-ECI, the actual accuracy of traditional machine learning models in predicting mental state classification is statistically analyzed. , and its calculation formula is: ;in: Indicates the number of correctly classified samples in the high ECI interval, Indicates the total number of samples in the high ECI interval.

[0097] Compare the classification accuracy in the high ECI range with the preset threshold θ:

[0098] like ≥θ, the model is considered to have good classification ability under complex emotion evolution;

[0099] like <θ, it is determined that the model is insufficient in capturing the contextual features of temporal psychological data.

[0100] When it is determined that the model accuracy is insufficient, the amount of data input that reflects the time series characteristics is dynamically adjusted, and the adjusted traditional machine learning model is iteratively trained and evaluated, specifically including:

[0101] Determine whether the classification accuracy of the traditional machine learning model within the high emotional coherence index range is below the preset threshold. If so, the model is judged to be insufficiently capable of capturing temporal features and enters the dynamic adjustment process.

[0102] The Dynamic Time Warping (DTW) algorithm is used as the basis for dynamic adjustment. DTW can nonlinearly align two time series and measure their similarity, even if they change at different speeds.

[0103] DTW calculation is applied to each student's continuous emotion score sequence or text sentiment tendency sequence. The distance feature based on DTW reflects the overall shape and evolution pattern of the student's emotion change trajectory.

[0104] Dynamically adjust input features based on the degree of change of DTW features:

[0105] If the DTW distance is greater than the preset distance threshold (indicating a dramatic change in sentiment), increase the data input dimension, such as introducing a longer time window of sentiment sequences;

[0106] If the DTW distance is less than or equal to the preset distance threshold (indicating that the emotion changes smoothly), the input amount is appropriately reduced to retain the core change segment and improve data processing efficiency.

[0107] The new time series features adjusted based on DTW are combined with the original static features (such as single-point sentiment score and text similarity) to form a new training dataset.

[0108] Retrain traditional machine learning models (such as SVM and random forest) using the updated dataset. Perform hyperparameter optimization during training, such as adjusting the regularization coefficient and decision tree depth.

[0109] Re-evaluate the classification accuracy of the adjusted model within the validation set and the high sentiment coherence index range. If it meets or exceeds the preset accuracy threshold, the optimization is complete. Otherwise, continue adjusting the feature input or switch to another deep time series model (such as LSTM).

[0110] This embodiment provides an intelligent screening method for identifying students' mental states. This method addresses the problem that traditional machine learning models have insufficient capture capabilities when processing psychological diary data with time series and context-dependent characteristics. By collecting student psychological diary data over continuous time periods, the emotional fluctuation amplitude characteristics and context-related characteristics are extracted, and an emotional coherence index is comprehensively calculated to represent the complexity of emotional evolution. Furthermore, the classification accuracy of the traditional model is evaluated within different emotional coherence index intervals. When the accuracy falls below a preset threshold in a high emotional coherence interval, a dynamic time warping (DTW) algorithm is used to dynamically adjust the data input amount reflecting the time series characteristics. The adjusted model is then iteratively trained and evaluated, thereby effectively improving the accuracy and robustness of the mental state recognition system for complex emotional evolution samples and achieving intelligent and accurate mental health screening.

[0111] Example 2, please refer to Figure 2 As shown, the student mental state recognition intelligent screening system described in this embodiment includes a data acquisition and preprocessing module, a feature extraction module, an emotion coherence calculation module, a mental state recognition and performance evaluation module, an accuracy detection and deficiency judgment module, and a dynamic adjustment and iterative optimization module;

[0112] Data collection and preprocessing module: obtains students' psychological diary data within a continuous time period, cleans the psychological diary data, and forms a standardized text data set;

[0113] Feature extraction module: extracting emotion fluctuation amplitude features and context relevance features from the text dataset;

[0114] Emotional coherence calculation module: calculates an emotional coherence index based on the emotional fluctuation amplitude feature and the context relevance feature, wherein the emotional coherence index is used to represent the complexity of emotional evolution;

[0115] Mental State Recognition and Performance Evaluation Module: This module classifies and predicts students' mental states based on the trained traditional machine learning model, and calculates the classification accuracy within different emotional coherence index ranges.

[0116] Accuracy detection and deficiency judgment module: determines whether the classification accuracy of the traditional machine learning model in the high emotional coherence index range is lower than a preset accuracy threshold. If so, it is determined that the model is insufficient in capturing the contextual features of the time-series psychological data;

[0117] Dynamic adjustment and iterative optimization module: When it is determined that the model accuracy is insufficient, the data input amount reflecting the time series characteristics is dynamically adjusted, and the adjusted traditional machine learning model is iteratively trained and evaluated.

[0118] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0119] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. An intelligent screening method for identifying students' psychological states, characterized by: include: Acquiring student psychological diary data within a continuous time period, and cleaning the psychological diary data to form a standardized text data set; Extracting emotion fluctuation amplitude features and context relevance features from the text dataset; Calculating an emotional coherence index based on the emotional fluctuation amplitude feature and the context relevance feature, wherein the emotional coherence index is used to indicate the complexity of emotional evolution; Based on the trained traditional machine learning model, the students' psychological states were classified and predicted, and the classification accuracy was statistically analyzed within different emotional coherence index ranges. Determining whether the classification accuracy of the traditional machine learning model within the high emotion coherence index range is lower than a preset accuracy threshold; if so, determining that the model is insufficient in capturing the contextual features of the time-series psychological data; When it is determined that the model accuracy is insufficient, the amount of data input that reflects the time series characteristics is dynamically adjusted, and the adjusted traditional machine learning model is iteratively trained and evaluated.

2. The intelligent screening method for identifying students' psychological states according to claim 1, characterized in that: The method for obtaining the emotional change amplitude value is: for each log text, apply the sentiment analysis model to generate the emotional score, which is recorded as , where t represents the time node, and the amplitude of a single emotion change is calculated based on the absolute value of the difference in emotion scores between two consecutive time nodes , the expression is: ; Calculate the average emotion change amplitude in a continuous time period. The calculation expression is: ; Where N is the total number of time nodes, and the calculation expression for the emotion change amplitude value ME is: .

3. The intelligent screening method for identifying students' psychological states according to claim 2, characterized in that: The method for obtaining text association similarity is to convert the log text of each time node into a fixed-dimensional vector representation through the embedding model, which is recorded as , calculate the cosine similarity between the text vectors of two consecutive time nodes as the single text association similarity , the calculation expression is: ;in: is the modulus of each vector, and the text association similarity is calculated: ; M is the total number of time points for obtaining text vectors, is the text association similarity, which is used to indicate the degree of context coherence.

4. The intelligent screening method for identifying students' psychological states according to claim 3, characterized in that: An emotional coherence index is calculated based on the emotional fluctuation amplitude feature and the context relevance feature. The emotional coherence index is used to indicate the complexity of emotional evolution, specifically including: The emotion change amplitude value and text association similarity are normalized so that they are both between [0, 1]. The emotion coherence index is calculated based on the normalized emotion change amplitude value and text association similarity.

5. The intelligent screening method for identifying students' psychological states according to claim 4, characterized in that: The byte comparison deviation index and the checksum comparison anomaly index are normalized so that they are both between [0, 1]. The accuracy score of the current message parsing is obtained by taking the weighted average of the normalized byte comparison deviation index and the checksum comparison anomaly index.

6. The intelligent screening method for identifying students' psychological states according to claim 5, characterized in that: Based on the trained traditional machine learning model, the students’ psychological states are classified and predicted, and the classification accuracy is statistically analyzed within different emotional coherence index ranges, including: Using a dataset with labeled mental state categories, a traditional machine learning model is used for model training. Input features include emotion scores, emotional tendencies, and behavioral pattern indicators, and the corresponding mental state category labels are output. Input the psychological diary data of the students to be tested into the trained model to automatically classify and predict their psychological states, and obtain the predicted labels for each student in each time period; For each student, the emotional coherence index (ECI) was calculated based on their psychological diary for consecutive time periods; All students were divided into different intervals according to their ECI values, including: low complexity interval: ECI < 0.3; medium complexity interval: 0.3 ≤ ECI < 0.6; high complexity interval: ECI ≥ 0.6; each interval contains several student samples; In each ECI interval, the classification accuracy of the model is calculated separately: Define whether the prediction of each sample is correct; Count the number of samples predicted correctly in each interval ; Count the total number of samples in each interval ; Calculate interval accuracy , the expression is: .

7. The intelligent screening method for identifying students' psychological states according to claim 6, characterized in that: Determine whether the classification accuracy of the traditional machine learning model within the high emotional coherence index range is lower than a preset accuracy threshold. If so, determine that the model is insufficient in capturing the contextual features of the time-series psychological data. Specifically, the following are included: The classification accuracy threshold θ is preset. Among all student samples, the sample set with the emotional coherence index ECI ≥ 0.6 is screened out and recorded as the high ECI interval dataset Dhigh-ECI; In Dhigh-ECI, the actual accuracy of traditional machine learning models in predicting mental state classification is statistically analyzed. , and its calculation formula is: ;in: Indicates the number of correctly classified samples in the high ECI interval, Indicates the total number of samples in the high ECI interval; Compare the classification accuracy in the high ECI interval with the preset threshold θ: ≥θ, the model is considered to have good classification ability under complex emotional evolution; if <θ, it is determined that the model is insufficient in capturing the contextual features of temporal psychological data.

8. The intelligent screening method for identifying students' psychological states according to claim 7, characterized in that: When it is determined that the model accuracy is insufficient, the amount of data input that reflects the characteristics of the time series is dynamically adjusted, including: Apply DTW calculation to each student's continuous emotion score sequence or text sentiment tendency sequence, and dynamically adjust the input features according to the degree of change of DTW features: If the DTW distance is greater than the preset distance threshold, the data input dimension is increased; If the DTW distance is less than or equal to the preset distance threshold, the input amount is reduced; Combine the new time series features adjusted based on DTW with the original static features to form a new training dataset; Retrain traditional machine learning models using updated datasets.

9. A student mental state recognition and intelligent screening system, used to implement the student mental state recognition and intelligent screening method according to any one of claims 1 to 8, characterized in that: It includes data acquisition and preprocessing module, feature extraction module, emotion coherence calculation module, psychological state recognition and performance evaluation module, accuracy detection and deficiency judgment module, and dynamic adjustment and iterative optimization module; Data collection and preprocessing module: obtains students' psychological diary data within a continuous time period, cleans the psychological diary data, and forms a standardized text data set; Feature extraction module: extracting emotion fluctuation amplitude features and context relevance features from the text dataset; Emotional coherence calculation module: calculates an emotional coherence index based on the emotional fluctuation amplitude feature and the context relevance feature, wherein the emotional coherence index is used to represent the complexity of emotional evolution; Mental State Recognition and Performance Evaluation Module: This module classifies and predicts students' mental states based on the trained traditional machine learning model, and calculates the classification accuracy within different emotional coherence index ranges. Accuracy detection and deficiency judgment module: determines whether the classification accuracy of the traditional machine learning model in the high emotional coherence index range is lower than a preset accuracy threshold. If so, it is determined that the model is insufficient in capturing the contextual features of the time-series psychological data; Dynamic adjustment and iterative optimization module: When it is determined that the model accuracy is insufficient, the data input amount reflecting the time series characteristics is dynamically adjusted, and the adjusted traditional machine learning model is iteratively trained and evaluated.

Images