Analysis System for Negative Emotional Representations in College Social Networks

Abstract

The present invention discloses a negative social emotion representation analysis system for colleges and universities, which relates to the technical field of social emotion analysis, and includes a data collection and preprocessing module, a feature extraction and anomaly detection module, an intelligent perception and negative emotion representation module, and an anomaly risk assessment and response module: The data collection and preprocessing module obtains emotion feature information in real time based on multi-dimensional data sources, and cleans and preprocesses the obtained emotion feature information. The present invention solves the problem of multicollinearity among multi-dimensional data features by calculating the data consistency offset coefficient and the strong correlation coefficient, avoiding unstable model regression coefficients. The system real-time identifies abnormal correlations and classifies and processes them in combination with risk thresholds to ensure robust model output. By continuously collecting data to establish an analysis set, it supports long-term emotion trend analysis and provides personalized responses according to the risk level, such as reminders, interventions or emergency measures, to achieve precise management and timely adjustment of students' emotional states.

Description

Technical Field

[0001] The present invention relates to the technical field of social emotion analysis, and particularly to a system for analyzing the representation of negative social emotions in colleges and universities. Background Art

[0002] The analysis of the representation of negative social emotions in colleges and universities refers to the reverse analysis and modeling of the emotional states of college students in social situations by using big data technology and theories of psychology and sociology. Its core goal is to deduce the deep psychological states and emotional fluctuations of students through information such as language, behavior, and emotional expressions in social platforms or interactions inside and outside the campus, especially those emotions not directly expressed on the surface. For example, students may show positive emotions (such as optimism and happiness) in public interactions, but reverse representation analysis can reveal underlying negative emotions (such as anxiety and loneliness), so as to understand their true emotional states more comprehensively and deeply. This analysis is of great significance in mental health intervention and campus management - schools can identify signals of long-term depression or social stress in a timely manner, discover potential psychological problems, and then formulate targeted psychological counseling plans to optimize the interactions among students. At the same time, as college students increasingly rely on social platforms to express themselves, this analysis provides data support for educational administrators, helping them master the trend of emotional changes, formulate preventive measures, and create a healthy and harmonious campus environment.

[0003] In specific operations, the system identifies explicit emotions (such as happiness, anger, and sadness) through sentiment analysis algorithms, and combines psychological models (such as depression tendency detection and loneliness index) to infer potential implicit emotions and inner states. Through natural language processing technology (NLP), the system extracts emotional features from text and behavioral data, such as emotion words, semantic structures, and emotional change trajectories. At the same time, behavioral data (such as like frequency and participation in activities) is combined with time series models to reveal the dynamic fluctuations of emotional states. Then, the system uses a multimodal fusion method to integrate features such as language, behavior, and emotional trajectories into a unified feature vector, and analyzes their mutual relationships and change trends. Combining with psychological models, the system can identify potential psychological risks and give early warnings, such as inferring whether a student may have a tendency of depression or loneliness. Finally, the analysis results provide accurate data support for schools, helping them take effective intervention measures before problems occur and improve the efficiency of students' mental health management.

[0004] The prior art has the following deficiencies:

[0005] In the analysis of negative social emotions in colleges and universities, multi-dimensional data (such as language, behavior, and emotional trajectories) are input into a fusion model after feature extraction to form a unified feature vector. However, when there is a strong correlation between the features of different data sources (such as a high correlation between emotional trajectories and language emotion scores), it will cause the problem of multicollinearity. Multicollinearity will make the regression coefficients or weights of the model unstable, making it difficult for the model to accurately judge the independent contributions of each feature to emotion judgment. Even if there is a slight change in the data input, the parameters may fluctuate violently. This instability will cause the model to output very different emotion state predictions at different times or datasets, resulting in uncontrollable results and making it difficult to conduct long-term monitoring and trend analysis.

[0006] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure. Therefore, it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] The object of the present invention is to provide a system for analyzing negative social emotions in colleges and universities. By calculating the data consistency offset coefficient and the strong correlation coefficient, it effectively solves the problem of multicollinearity caused by the strong correlation between multi-dimensional data features, and avoids the risk of instability of the model regression coefficients and weights. The system can identify abnormal correlations in real time, and classify and process emotion feature information in combination with risk thresholds to ensure the robustness and accuracy of the model output. By continuously collecting data to establish an analysis set, the system supports long-term emotion trend analysis and provides personalized responses based on risk levels, such as reminders, interventions, or emergency measures. This dynamic monitoring and response mechanism ensures the accurate management and timely adjustment of students' emotion states to solve the problems in the above background art.

[0008] To achieve the above object, the present invention provides the following technical solution: A system for analyzing negative social emotions in colleges and universities, including a data collection and preprocessing module, a feature extraction and anomaly detection module, an intelligent perception and negative emotion representation module, and an anomaly risk assessment and response module:

[0009] The data collection and preprocessing module obtains emotion feature information in real time based on multi-dimensional data sources, cleans and preprocesses the obtained emotion feature information to ensure that different emotion feature information is analyzed on the same dimension;

[0010] The feature extraction and anomaly detection module extracts key features from the preprocessed emotion feature information, and then performs anomaly analysis on the extracted feature information to detect and mark the atypical associations between emotion feature information;

[0011] The intelligent perception and negative emotion representation module deploys a pre-trained machine learning model in the emotion monitoring system. Through the intelligent perception ability of the machine learning model, it dynamically perceives the subtle changes in the feature data, identifies potential abnormal fluctuations, and maps them to the negative emotion representation results;

[0012] The abnormal risk assessment and response module further analyzes the identified potential abnormal fluctuations, determines the risk level of the potential abnormal fluctuations, and formulates different response measures for different potential abnormal fluctuation risk levels.

[0013] Preferably, the emotion feature information includes language emotion polarity information and circadian emotion variation information. The language emotion polarity information refers to judging the emotional direction and the degree of emotional intensity by analyzing the words, grammar, and sentence patterns in the language text or speech expression. The circadian emotion variation information is used to measure the change range and pattern of an individual's emotional state at different time periods within a day.

[0014] Preferably, after obtaining the language emotion polarity information and circadian emotion variation information in the emotion feature information, an abnormal analysis is performed on the language emotion polarity information to generate a language emotion polarity fluctuation index, and an abnormal analysis is performed on the circadian emotion variation information to generate a circadian emotion variation index. The atypical correlation between the obtained multi-dimensional data is identified through the generated language emotion polarity fluctuation index and circadian emotion variation index.

[0015] Preferably, the language emotion polarity fluctuation index and circadian emotion variation index generated after abnormal analysis obtained through multi-dimensional data sources are input into a pre-trained machine learning model. A data consistency offset coefficient is generated through the machine learning model, and the obtained emotion feature information is intelligently evaluated through the data consistency offset coefficient.

[0016] Preferably, the generated data consistency offset coefficient is compared and analyzed with a pre-set data consistency offset coefficient reference threshold to identify the strong correlation between the obtained emotion characteristic information. The analysis results are as follows:

[0017] If the data consistency offset coefficient is greater than or equal to the pre-set data consistency offset coefficient reference threshold, a potential abnormal correlation is generated, indicating that there is a strong correlation between the obtained emotion characteristic information;

[0018] If the data consistency offset coefficient is less than the pre-set data consistency offset coefficient reference threshold, an emotion characteristic independence is generated, indicating that there is no strong correlation between the obtained emotion characteristic information.

[0019] Preferably, when there is a strong correlation between the obtained emotion characteristic information and a potential abnormal correlation is generated, continuously obtain the data consistency offset coefficient generated after abnormal analysis of the emotion characteristic information, establish an analysis set, and compare the data consistency offset coefficient in the analysis set with the first-level reference threshold, the second-level reference threshold, and the data consistency offset coefficient reference threshold. Among them, the second-level reference threshold is greater than the first-level reference threshold, and the first-level reference threshold is greater than the data consistency offset coefficient reference threshold. Compare and analyze the data consistency offset coefficient with the second-level reference threshold, the first-level reference threshold, and the data consistency offset coefficient reference threshold. Calibrate the number of data consistency offset coefficients that are less than the first-level reference threshold and greater than or equal to the data consistency offset coefficient reference threshold as Fa, calibrate the number of data consistency offset coefficients that are less than the second-level reference threshold and greater than or equal to the first-level reference threshold as Fb, and calibrate the number of data consistency offset coefficients that are greater than or equal to the second-level reference threshold as Fc;

[0020] Comprehensively analyze Fa, Fb, and Fc to generate a strong correlation coefficient SCC. The formula is as follows:

[0021]

[0022] , where h1, h2, and h3 are the preset proportionality coefficients of Fa, Fb, and Fc respectively, and Fa, Fb, and Fc are all greater than 0.

[0023] Preferably, compare and analyze the generated strong correlation coefficient with the preset first strong correlation coefficient reference threshold and the second strong correlation coefficient reference threshold. The analysis results are as follows;

[0024] If the strong correlation coefficient is less than the first strong correlation coefficient reference threshold, then further divide the potential abnormal correlation into a low-risk strong correlation anomaly;

[0025] If the strong correlation coefficient is greater than or equal to the first strong correlation coefficient reference threshold and less than the second strong correlation coefficient reference threshold, then further divide the potential abnormal correlation into a medium-risk strong correlation anomaly:

[0026] If the strong correlation coefficient is greater than or equal to the second strong correlation coefficient reference threshold, then further divide the potential abnormal correlation into a high-risk strong correlation anomaly.

[0027] Preferably, the logic for abnormal analysis of language emotion polarity information to generate a language emotion polarity fluctuation index is as follows:

[0028] Under the detection window, for a given language input, extract the emotion polarity score S through an emotion analysis model i, and convert it into time - series data. The input language text time - series data is as follows: T = {t i} = {t1, t2, t3, ……, t n}, where T represents the language text sequence collected within the detection window, t i represents the i - th language content at time point t, n represents the total number of texts collected within the detection window, and the output sentiment polarity sequence data is as follows: S = {S i} = {S1, S2, ……, S m}, where S represents the sentiment polarity sequence, S i represents the sentiment polarity score of the i - th language content at time point t, m represents the total number of language texts within the detection window, that is, how many language text data are analyzed in total, reflecting how many sentiment records are processed within the detection window;

[0029] Calculate the sentiment polarity change rate at each time point on the sentiment polarity sequence S, capture the sentiment polarity change trend through high - order differences, and enhance the sensitivity to mutations. The calculation formula is as follows:

[0030]

[0031] , where G t represents the sentiment polarity change gradient at time point t, is the second - order difference, measuring the acceleration of the sentiment polarity change and capturing sharp fluctuations, is the first - order difference, capturing the sentiment polarity change rate at consecutive time points, and α is a balance coefficient used to adjust the influence of the first - order and second - order gradients;

[0032] Introduce a non - linear weight function to emphasize the sentiment polarity anomaly of each sentiment polarity change gradient G t . The expression is as follows: where W(G t ) is the non - linear weight parameter, giving higher influence to sharp sentiment polarity changes and reducing the interference of small fluctuations, β is the weight amplification coefficient used to control the non - linear amplification degree, and e represents the natural base;

[0033] Cumulate the non - linear weights at all time points to generate the language sentiment polarity fluctuation index. The expression is as follows:

[0034]

[0035] , where SPFI represents the language sentiment polarity fluctuation index, γ is the non - linear scaling coefficient used to control the influence of the polarity score in the fluctuation index, and max(|S|) is the maximum absolute value in the sentiment polarity sequence S.

[0036] Preferably, the abnormal analysis of the circadian mood variation information and the generation of the circadian mood variation index are based on the following logic:

[0037] Under the detection window, continuously collect the mood data of the target individual, and label the collected mood data as E t , E t is the mood state value of the detected individual at time point t;

[0038] Since the mood state fluctuates over time, we need to calculate the instantaneous change rate of the time series to capture the amplitude of mood changes. Using the form of the derivative, define the mood change rate function, and the calculation expression is as follows:

[0039]

[0040] , where ΔE(t) represents the instantaneous mood change rate, reflecting the mood fluctuation amplitude at time point t, and f circ (t) is the circadian rhythm weight function, used to consider the periodic influence of day and night,

[0041] where, 2π is the factor used to map the cycle of one day to the cycle of the trigonometric function;

[0042] In the instantaneous mood change rate ΔE(t), introduce a non-linear fluctuation intensity function to detect abnormal mood fluctuations, which is used to quantify the abnormal fluctuation amplitude of mood, and the expression is as follows:

[0043]

[0044] , where F anom (t) represents the abnormal mood fluctuation intensity function, t0 represents the starting time of mood monitoring, t1 represents the ending time of mood monitoring, θ is the non-linear adjustment parameter, used to control the sensitivity of mood fluctuations, λ is the adjustment weight, used to control the influence of mood state on abnormal fluctuations, and g(E(t)) is the mood state adjustment function, where g(E(t)) = exp(-σE(t)), where σ is the parameter controlling the attenuation rate;

[0045] Based on the time-weighted result of the abnormal mood fluctuation intensity function and the circadian rhythm, generate the circadian mood variation index, and the calculation expression is as follows:

[0046]

[0047] , where DEVI represents the circadian mood variation index, and k is the circadian rhythm weight adjustment factor.

[0048] In the above technical solution, the technical effects and advantages provided by the present invention:

[0049] The present invention effectively solves the problem of multicollinearity caused by strong correlations between multi-dimensional data features by calculating the data consistency offset coefficient and the strong correlation coefficient. This collinearity problem often leads to unstable regression coefficients and weights of the model, affecting the accurate judgment of the independent contribution of emotional features to the emotional state. By real-time identifying potential abnormal correlations and combining with the set risk threshold, the system can accurately classify and process emotional feature information when multicollinearity occurs, ensuring the output stability of the model. This mechanism improves the accuracy of emotional state prediction, avoids the volatility of prediction results caused by data correlations, ensures the robustness of the model across different time periods and datasets, and thus enhances the accuracy and reliability of emotional monitoring.

[0050] The present invention continuously collects emotional feature information, dynamically generates the data consistency offset coefficient, and establishes an analysis set to support the analysis of long-term emotional change trends. By comparing the data in the analysis set with the set threshold, the system can identify the risk level of the emotional feature information and provide personalized response strategies based on this. For example, low-risk anomalies may only require regular emotional management reminders, medium risks will trigger psychological counseling interventions, and high risks will activate an emergency response mechanism. This multi-level risk analysis and response strategy enables the system to continuously and dynamically monitor the emotional state, support the long-term trend analysis and refined management of emotional features, and ensure that the emotional state of students can be managed and regulated in a timely and effective manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] To more clearly illustrate the technical solutions in the embodiments of the present application or in the prior art, the following will briefly introduce the accompanying drawings required in the embodiments. Obviously, the accompanying drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0052] Figure 1 It is a schematic diagram of the modules of the college social negative emotion characterization analysis system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] Now, the exemplary embodiments will be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0054] The present invention provides a college social negative emotion characterization analysis system as Figure 1 shown, including a data collection and preprocessing module, a feature extraction and anomaly detection module, an intelligent perception and negative emotion characterization module, and an anomaly risk assessment and response module:

[0055] A data acquisition and preprocessing module that obtains emotional feature information in real time based on multi-dimensional data sources, cleans and preprocesses the obtained emotional feature information to ensure that different emotional feature information is analyzed on the same dimension;

[0056] The emotional feature information includes language emotional polarity information and circadian emotional variation information. The language emotional polarity information refers to judging the emotional direction (positive, neutral or negative) and the degree of emotional intensity by analyzing the vocabulary, grammar and sentence patterns in language texts or speech expressions. The circadian emotional variation information is used to measure the change range and pattern of an individual's emotional state at different time periods within a day (such as morning, afternoon, night);

[0057] The multi-dimensional data sources cover various feature information of an individual in social situations, mainly including the following categories:

[0058] Language feature data: such as language emotional polarity information, speech rate, intonation and emotional word frequency in dialogue texts, comments, and speech records.

[0059] Behavioral feature data: including micro-expressions, facial expression changes, body movements and interaction behavior activity levels.

[0060] Physiological feature data: collected through wearable devices, such as heart rate variability, skin conductance response, respiratory rate and sleep quality.

[0061] Time series emotional data: such as circadian emotional variation information, emotional stability and behavioral rhythm characteristics.

[0062] These data sources reflect an individual's psychological state and social behavior from different angles, providing comprehensive support for the analysis of negative emotion representation

[0063] A feature extraction and anomaly detection module that extracts key features from the preprocessed emotional feature information, and then performs anomaly analysis on the extracted feature information to detect and mark atypical associations between emotional feature information;

[0064] After obtaining the language emotional polarity information and circadian emotional variation information in the emotional feature information, perform anomaly analysis on the language emotional polarity information to generate a language emotional polarity fluctuation index, and perform anomaly analysis on the circadian emotional variation information to generate a circadian emotional variation index. Identify atypical associations between the obtained multi-dimensional data through the generated language emotional polarity fluctuation index and circadian emotional variation index;

[0065] Atypical associations between emotional characteristic information refer to abnormal or unexpected relationships existing between different data types, which do not conform to conventional patterns or statistical laws. For example, in sentiment analysis, we usually expect positive linguistic emotional expressions to be associated with high social activity. However, if there is a coexistence of positive language and low interaction behavior, this abnormal combination is an atypical association. Atypical associations often reveal an individual's complex psychological state, such as emotion concealment, behavioral inhibition, or a disorder between physiological state and psychological manifestation.

[0066] The atypical associations existing between emotional characteristic information are as follows:

[0067] 1. Association between positive linguistic emotional polarity and low social interaction: An individual's language expression is very positive, but shows social withdrawal in behavior, which may indicate emotion concealment or potential social anxiety.

[0068] 2. Association between circadian emotional variation and sufficient sleep: Although the sleep duration is sufficient, the emotional variation is large, which may indicate a disruption of the biological rhythm or emotional instability caused by psychological stress.

[0069] 3. Association between high speech rate and low emotional intensity: The speech rate is accelerated but the language lacks emotional expression, which may imply an anxious state or emotional numbness under psychological stress.

[0070] 4. Association between high heart rate and calm emotional expression: The heart rate rises significantly but the external emotional expression is calm, indicating that the individual may be suppressing emotions or has experienced a high-stress situation.

[0071] These atypical associations reveal the complex relationships between emotion, behavior, and physiological state, and require comprehensive analysis to avoid misjudgment caused by simple linear inference.

[0072] During the data fusion process, if there are significant changes in the language sentiment polarity, it means that there may be strong correlations among the preprocessed emotion feature information, which can easily lead to the problem of multicollinearity. The reason for the occurrence of multicollinearity is that there is a high degree of linear correlation among the features input into the model, making it impossible for the model to effectively distinguish the independent contributions of each feature. For example, there may be strong internal correlations between language sentiment polarity and features such as diurnal mood variation and speech rate variation: when an individual's mood fluctuates greatly over time, the emotional expression in language will tend to be extreme, or the speech rate will increase in a highly anxious state, and at the same time, the language sentiment polarity will also increase significantly. This strong correlation will cause instability in the coefficients of the model in regression analysis, and even cause drastic fluctuations in the model output when the data changes slightly. In addition, if the model assigns high weights to these redundant features, it will lead to a decrease in the interpretability of the model, making it difficult to clarify the independent contributions of each feature to the final emotion judgment. Especially in the application scenario of emotion monitoring, this instability will bring prediction biases, resulting in incorrect trend judgments or monitoring misjudgments.

[0073] The logic for performing anomaly analysis on language sentiment polarity information and generating the language sentiment polarity fluctuation index is as follows:

[0074] Under the detection window, for a given language input (such as a dialogue or text), extract the sentiment polarity score S through the sentiment analysis model i , and convert it into time series data. The input language text time series data is as follows: T = {t i} = {t1, t2, t3,..., t n}, where T represents the language text sequence collected within the detection window, and t i represents the i-th language content (such as a comment or a dialogue statement) at time point t. i ranges from 1 to n, indicating the order of the text in the time series, and n represents the total number of language texts collected within the detection window, that is, how many language content expressions are included in the analysis in the time series. Each language content corresponds to a time point. The output sentiment polarity sequence data is as follows: S = {S i} = {S1, S2,..., S m}, where S represents the sentiment polarity sequence, and S i represents the sentiment polarity score of the i-th language content at time point t. m represents the total number of language texts within the detection window, that is, how many language text data are analyzed in total, reflecting how many emotion records are processed within the detection window;

[0075] In emotion analysis, the language text T consists of a series of text units (such as sentences, phrases, or dialogues). These texts cannot be directly used to calculate or analyze the sentiment polarity fluctuation, but need to be processed through the sentiment analysis model to extract the numerical features related to emotion, that is, the sentiment polarity score Si 。

[0076] Processed by the sentiment analysis model to extract numerical features related to sentiment, namely the sentiment polarity score S i The steps are as follows:

[0077] First, the original text T = {t i} = {t1, t2, t3,..., t n} needs to be cleaned and normalized to ensure efficient processing by the sentiment analysis model.

[0078] After text preprocessing, each text unit t i is passed as input to the sentiment analysis model. The model determines the sentiment tendency of the text based on the vocabulary and context and generates a sentiment polarity score S i 。

[0079] For each text unit t i the analysis result is a sentiment polarity score S i indicating the sentiment tendency and intensity of the text.

[0080] Finally, the entire language text T is converted into a sentiment polarity sequence S = {S i} = {S1, S2,..., S m}, where each S i corresponds to the sentiment polarity score of a text unit t i .

[0081] In sentiment analysis and mood fluctuation analysis, the given language input refers to the text or language data used for model analysis. These inputs come from various social, conversational, or recorded environments and are used to extract sentiment information to reflect changes in mood states. The language input is the core data source of the sentiment analysis system and will be converted into numerical features, such as the sentiment polarity score Si, for further model analysis.

[0082] Sources of the given language input

[0083] 1. Social platform data: Short text data such as user comments, status updates, Weibo posts, forum posts, etc.

[0084] 2. Conversation records: Chat records of instant messaging tools (such as WeChat, Slack), or conversation texts of customer service.

[0085] 3. Speeches and interviews: Texts transcribed from recordings, such as meeting minutes, interview contents.

[0086] 4. Written texts: Words in news reports, blog articles, email bodies, research papers.

[0087] 5. Spoken-to-written data: The dialogue content generated by speech transcription technology (such as the conversion of recorded speech to text in voice assistants).

[0088] Features of the given language input

[0089] 1. Different degrees of structuring:

[0090] Some language inputs are structured (such as customer service dialogues with a clear question-and-answer order), while some are unstructured (such as user comments on social platforms).

[0091] 2. Context dependence:

[0092] Certain language inputs (such as chat conversations) require considering context information to accurately identify the emotional tendency. For example, "It's amazing" may carry sarcastic or genuine positive emotions depending on the context.

[0093] 3. Diverse expressions:

[0094] Language inputs may contain various expressions such as emojis, abbreviations, slang, irony, etc., all of which require the model to appropriately identify and understand.

[0095] The role of language input in sentiment analysis

[0096] 1. Data source: Language input is the basis for the model to extract emotional features. By analyzing these inputs, a quantitative sentiment polarity score S is generated i .

[0097] 2. Capturing emotional fluctuations: Organizing these input data in a time series can capture the dynamic change trends of emotions.

[0098] 3. Support for quantitative analysis: Through the extracted polarity scores, further calculate indicators such as the emotional fluctuation index to reveal the changes in emotional states and the correlations between features.

[0099] The given language input refers to the text or dialogue data used by the model for sentiment analysis, and these data reflect the emotional expressions of individuals in different scenarios. After preprocessing and emotion extraction, the language input will be converted into a sentiment polarity score, providing a basis for further emotion monitoring and data fusion.

[0100] Calculate the rate of change of sentiment polarity at each time point on the sentiment polarity sequence S, capture the trend of sentiment polarity change through high-order differences, and enhance the sensitivity to mutations. The calculation expression is as follows:

[0101]

[0102] , where G tRepresents the gradient of the change in sentiment polarity at time point t, is the second-order difference, which measures the acceleration of the change in sentiment polarity and captures sharp fluctuations, is the first-order difference, which captures the rate of change in sentiment polarity at consecutive time points. α is a balance coefficient used to adjust the influence of the first-order and second-order gradients;

[0103] Introduce a non-linear weight function to emphasize the sentiment polarity anomaly of each sentiment polarity change gradient G t as follows: where W(G t ) is the non-linear weight parameter, which assigns higher influence to sharp changes in sentiment polarity and reduces the interference of small fluctuations. β is the weight amplification coefficient used to control the degree of non-linear amplification, and e represents the natural base;

[0104] Assigning higher influence to sharp changes in sentiment polarity and reducing the interference of small fluctuations means that by introducing a non-linear weight function, the model pays more attention to those significant and sudden sentiment fluctuations when processing sentiment data, while ignoring or weakening those frequent but small-amplitude sentiment changes. The core of this method lies in: not all sentiment fluctuations contribute equally to the analysis results in the process of sentiment analysis. Sharp changes in sentiment polarity often mean potential mood anomalies (such as sudden outbursts of anger or anxiety), so higher weights need to be given to highlight their importance. Small-amplitude sentiment fluctuations may be noise or normal mood fluctuations and should not have too much impact on the overall analysis results. Through a non-linear weight function (such as an exponential function), the model can amplify large polarity changes and automatically filter out the interference of small fluctuations, thereby improving the robustness of the analysis and the ability to identify mood anomalies.

[0105] Accumulate the non-linear weights at all time points to generate the language sentiment polarity fluctuation index, as follows:

[0106]

[0107] , where SPFI represents the language sentiment polarity fluctuation index, γ is the non-linear scaling coefficient used to control the influence of the polarity score on the fluctuation index, and max(|S|) is the maximum absolute value in the sentiment polarity sequence S, which is used to normalize each sentiment polarity score S i ;

[0108] The expression for calculating the Language Sentiment Polarity Fluctuation Index indicates that a larger value of the Language Sentiment Polarity Fluctuation Index, generated after anomaly analysis of language sentiment polarity information, indicates significant fluctuations in language sentiment polarity information within the detection window. This indicates a high degree of strong correlation between emotional features (such as language sentiment, behavioral characteristics, and time series sentiment), which increases the probability of multicollinearity. This is because when the fluctuations of a particular feature are highly synchronized with those of other features, the model struggles to distinguish the independent contributions of each feature, leading to unstable weights and regression coefficients. Conversely, a smaller value of the Language Sentiment Polarity Fluctuation Index indicates lower correlations between features, greater feature independence, and a relatively lower risk of multicollinearity in the model.

[0109] During the data fusion process, changes in diurnal mood variability may indicate strong correlations between preprocessed mood feature information, leading to multicollinearity. The diurnal mood variability index reflects the magnitude of changes in individual mood over time and has a potential strong correlation with other emotional features (such as verbal emotional polarity, anxiety level, and sleep quality). For example, when an individual's mood fluctuates frequently, their verbal emotional polarity tends to show extremes, or decreased sleep quality may directly affect diurnal mood variability. The high correlation between these emotional features makes it difficult to distinguish their independent contributions in the fusion model, and the model's regression coefficients or weights become unstable, thus causing multicollinearity. Multicollinearity not only leads to inconsistent model performance across different datasets, but also may cause unreliable parameter estimation and difficult interpretation of results. Especially in long-term mood monitoring, it can make the model sensitive to small changes in the data, reducing its robustness.

[0110] The logic for performing anomaly analysis on diurnal mood variation information and generating a diurnal mood variation index is as follows:

[0111] Under the detection window, the emotional data of the target individual is continuously collected and the collected emotional data is calibrated as E t , E t It is the emotional state value of the detected individual at time point t;

[0112] Emotional data refers to information that reflects an individual's mental state and emotional expression, typically derived from multiple dimensions, including language, behavior, physiology, and the environment. This data can be structured or unstructured and plays a key role in emotion monitoring and analysis models. The following are the main categories of emotion data and detailed explanations:

[0113] 1. Language feature data

[0114] Sources: Text or audio from social media, chat logs, speeches, interviews, etc.

[0115] Emotional polarity score: Represents positive, negative, or neutral emotions in language expressions (e.g., calculated through an NLP model).

[0116] Emotion word frequency: The frequency of use of specific emotion words (such as "happy" and "angry").

[0117] Speech rate and intonation: An accelerated speech rate or changes in intonation may indicate anxiety or excitement.

[0118] Keyword extraction: Identifying key emotion themes (such as stress and achievement) through semantic analysis.

[0119] Whether it must be collected:

[0120] If language is the main channel of emotional expression, it is very important to collect language data, such as in social situations or dialogue analysis.

[0121] However, in some non-verbal environments (such as sports behavior or sleep monitoring), it can be selectively not collected.

[0122] 2. Behavioral characteristic data

[0123] Source: Micro-expression analysis, body language monitoring, interaction behavior recording.

[0124] Facial expression analysis: Capturing minute emotional reactions through expressions, such as surprise, anger, sadness, etc.

[0125] Behavioral activity: Reflects the degree of social participation and the frequency of behaviors, such as the initiative in conversation or the duration of participation.

[0126] Behavioral inhibition or avoidance: Observing whether an individual has avoidance behaviors or social withdrawal, revealing potential anxiety or depression.

[0127] Movement and posture data: Changes in body movements (such as gestures and walking frequency) can map emotional states.

[0128] Whether it must be collected:

[0129] In the study of non-verbal expressions (such as micro-expressions or body movements), behavioral data is particularly crucial.

[0130] However, in situations such as remote interviews or text analysis, behavioral data can be not collected.

[0131] 3. Physiological characteristic data

[0132] Source: Wearable devices, biosensors, electrocardiograms, etc.

[0133] Heart rate variability (HRV): High heart rate variability is usually associated with a relaxed state, while low variability may indicate a stressed state.

[0134] Electrodermal Activity (EDA): Changes in skin conductivity can reflect the level of emotional arousal.

[0135] Respiratory Rate: Faster breathing may be associated with anxiety or nervousness.

[0136] Sleep Quality: Analyze emotional stability and health status through sleep duration and stages.

[0137] Whether collection is necessary:

[0138] Physiological data is very important in high-precision emotion monitoring, such as anxiety and stress detection. However, the collection of these data requires equipment support, so it can be selectively collected according to the research purpose.

[0139] 4. Time-Series Emotion Data

[0140] Source: Emotion monitoring logs, emotion fluctuation data within a time period.

[0141] Diurnal Emotion Variation Information (DEVI): Reflects the amplitude and pattern of emotional fluctuations throughout the day.

[0142] Emotion State Labels: Record the emotion state at different time points (such as scoring every hour).

[0143] Emotion Trends and Periodicity: Identify long-term and short-term trends and periodic changes in emotions.

[0144] Whether collection is necessary:

[0145] Time-series data is crucial when analyzing emotion change patterns and long-term monitoring. If it is only for short-term emotion analysis, it can be not collected.

[0146] 5. Environmental Data

[0147] Source: Environmental sensors or geographical location data.

[0148] Environmental Volume and Brightness: High noise or overly bright light may cause emotional discomfort.

[0149] Temperature and Humidity: Extreme weather may affect the emotional state.

[0150] Geographical Location: Changes in location may be related to emotional fluctuations (such as being more relaxed at home and more tense at the office).

[0151] Whether collection is necessary:

[0152] Environmental data has an indirect impact on emotions. If emotion changes are related to the environment, it should be collected. However, in situations where mainly language or physiological data is analyzed, environmental data can be ignored.

[0153] It is not necessary to collect all data. The collection of emotional data needs to be determined in combination with specific application scenarios and analysis objectives. For example:

[0154] If the focus is on analyzing an individual's social behavior, language and behavioral data are particularly crucial, and physiological or environmental data can be ignored.

[0155] If the research is on stress or anxiety states, physiological data is necessary, but language data may not be required.

[0156] If studying long-term emotional change trends, time series emotional data and circadian rhythm information need to be collected.

[0157] Emotional states fluctuate over time. Therefore, we need to calculate the instantaneous change rate of the time series to capture the amplitude of emotional changes. In the form of a derivative, define the emotional change rate function, and the calculation expression is as follows:

[0158]

[0159] , where ΔE(t) represents the instantaneous emotional change rate, reflecting the amplitude of emotional fluctuations at time point t, and f circ (t) is the circadian rhythm weight function, used to consider the periodic influence of day and night (such as biological rhythms). Among them, 2π is a factor used to map the cycle of a day to the cycle of a trigonometric function, which ensures that f circ (t) completes a full cycle (i.e., from 0 to 24 hours), corresponding to the day-night cycle in biological rhythms;

[0160] In the instantaneous emotional change rate ΔE(t), introduce a non-linear fluctuation intensity function to detect abnormal emotional fluctuations, used to quantify the amplitude of abnormal emotional fluctuations. The expression is as follows:

[0161]

[0162] , where F anom (t) represents the abnormal emotional fluctuation intensity function, t0 represents the starting time of emotional monitoring, t1 represents the ending time of emotional monitoring, θ is a non-linear adjustment parameter, used to control the sensitivity of emotional fluctuations, λ is an adjustment weight, used to control the influence of emotional states on abnormal fluctuations, and g(E(t)) is an emotional state adjustment function, used to dynamically adjust the influence of emotional states on abnormal fluctuation detection during the calculation process. Among them, g(E(t)) = exp(-σE(t)), where σ is a parameter that controls the attenuation rate;

[0163] Based on the time-weighted result of the abnormal emotional fluctuation intensity function and the circadian rhythm, generate the circadian emotional variation index. The calculation expression is as follows:

[0164]

[0165] , where DEVI represents the diurnal mood variation index, and κ is the diurnal rhythm weight adjustment factor, which determines the contribution of the diurnal rhythm to the variation index;

[0166] It can be seen from the calculation expression of the diurnal mood variation index that the larger the value of the diurnal mood variation index generated after abnormal analysis of the diurnal mood variation information, the stronger the correlation degree between the mood characteristics within the detection window usually indicates. In this case, the strong correlation between the diurnal mood variation information and other mood characteristics (such as language emotional polarity, behavior inhibition, anxiety level, etc.) may be enhanced, increasing the probability of the problem of multicollinearity occurring in the model. When multicollinearity exists, it is difficult for the model to accurately identify the independent contribution of each feature to the result, and the regression coefficients may be unstable, resulting in fluctuations in the prediction results. While when the value of the diurnal mood variation index is smaller, it means that the correlation between the mood characteristics is weaker, and the contribution of each feature to the model is more independent. Therefore, the probability of causing the problem of multicollinearity is also lower.

[0167] The intelligent perception and negative mood representation module deploys a pre-trained machine learning model in the mood monitoring system. Through the intelligent perception ability of the machine learning model, it dynamically perceives the subtle changes in the feature data, identifies potential abnormal fluctuations, and maps them to the negative mood representation results;

[0168] The language emotional polarity fluctuation index SPFI and the diurnal mood variation index DEVI generated after abnormal analysis obtained through multi-dimensional data sources are input into the pre-trained machine learning model. Through the machine learning model, the data consistency offset coefficient DCDC is generated, and the obtained mood feature information is intelligently evaluated through the data consistency offset coefficient DCDC;

[0169] The machine learning model is not limited here. Any model that can generate the data consistency offset coefficient DCDC from the language emotional polarity fluctuation index SPFI and the diurnal mood variation index DEVI can be used. To implement the technical solution of the present invention, the present invention provides a specific implementation manner;

[0170] The calculation formula for generating the data consistency offset coefficient DCDC is as follows:

[0171] <0-

[0172] , where d1 and d2 are the preset proportionality coefficients of the language emotional polarity fluctuation index SPEI and the diurnal mood variation index DEVI, and both d1 and d2 are greater than 0.

[0173] It can be seen from the calculation expression of the data consistency offset coefficient that the larger the performance value of the language emotion polarity fluctuation index generated after abnormal analysis of the language emotion polarity information, and the larger the performance value of the day-night emotion variation index generated after abnormal analysis of the day-night emotion variation information, that is, the larger the performance value of the data consistency offset coefficient, it indicates that the strong correlation between the obtained emotion feature information is relatively high, thus increasing the probability of the occurrence of the multicollinearity problem. On the contrary, it indicates that the strong correlation between the obtained emotion feature information is relatively low, and the probability of the occurrence of the multicollinearity problem is smaller;

[0174] Compare and analyze the generated data consistency offset coefficient with the pre-set reference threshold of the data consistency offset coefficient to identify the strong correlation between the obtained emotion characteristic information. The analysis results are as follows:

[0175] If the data consistency offset coefficient is greater than or equal to the pre-set reference threshold of the data consistency offset coefficient, potential abnormal correlation is generated, indicating that there is a strong correlation between the obtained emotion characteristic information;

[0176] If the data consistency offset coefficient is less than the pre-set reference threshold of the data consistency offset coefficient, emotion characteristic independence is generated, indicating that there is no strong correlation between the obtained emotion characteristic information;

[0177] The abnormal risk assessment and response module further analyzes the identified potential abnormal fluctuations, judges the risk level of the potential abnormal fluctuations, and formulates different response measures for different potential abnormal fluctuation risk levels.

[0178] When there is a strong correlation between the obtained emotion characteristic information and potential abnormal correlation is generated, continuously obtain the data consistency offset coefficient generated after abnormal analysis of the emotion characteristic information, establish an analysis set, and compare the data consistency offset coefficient in the analysis set with the first-level reference threshold, the second-level reference threshold, and the reference threshold of the data consistency offset coefficient. Among them, the second-level reference threshold is greater than the first-level reference threshold, and the first-level reference threshold is greater than the reference threshold of the data consistency offset coefficient. Compare and analyze the data consistency offset coefficient with the second-level reference threshold, the first-level reference threshold, and the reference threshold of the data consistency offset coefficient. Mark the number of data consistency offset coefficients that are less than the first-level reference threshold and greater than or equal to the reference threshold of the data consistency offset coefficient as Fa, mark the number of data consistency offset coefficients that are less than the second-level reference threshold and greater than or equal to the first-level reference threshold as Fb, and mark the number of data consistency offset coefficients that are greater than or equal to the second-level reference threshold as Fc.

[0179] Comprehensively analyze Fa, Fb, and Fc to generate the strong correlation coefficient SCC. The formula is as follows:

[0180]

[0181] where h1, h2, and h3 are the preset proportionality coefficients of Fa, Fb, and Fc respectively, and Fa, Fb, and Fc are all greater than 0.

[0182] It can be seen from the strong correlation coefficient calculation expression that the larger the value of the strong correlation coefficient performance, the greater the probability that the collected emotional characteristic information shows strong correlation, and the smaller the value of the strong correlation coefficient performance, the smaller the probability that the collected emotional characteristic information shows strong correlation.

[0183] Compare and analyze the generated strong correlation coefficient with the preset first strong correlation coefficient reference threshold and second strong correlation coefficient reference threshold, and the analysis results are as follows;

[0184] If the strong correlation coefficient is less than the first strong correlation coefficient reference threshold, the potential abnormal correlation is further divided into low-risk strong correlation anomalies;

[0185] If the strong correlation coefficient is greater than or equal to the first strong correlation coefficient reference threshold and less than the second strong correlation coefficient reference threshold, the potential abnormal correlation is further divided into medium-risk strong correlation anomalies:

[0186] If the strong correlation coefficient is greater than or equal to the second strong correlation coefficient reference threshold, the potential abnormal correlation is further divided into high-risk strong correlation anomalies;

[0187] When a low-risk strong correlation anomaly is detected, it indicates that although the data consistency offset coefficient exceeds the reference threshold, the impact is small, and the interference of multicollinearity on the stability of the model and the analysis results is limited. At this time, the system will send a low-level warning prompt to remind the data analyst to pay attention to potential risks, but there is no need for immediate intervention. The sampling frequency should be increased to continuously monitor the data, and re-analysis of the data should be carried out when necessary to ensure reasonable feature selection of the model.

[0188] When a medium-risk strong correlation anomaly is detected, it means that the data offset and feature correlation have caused greater interference to the model, and the model needs to be further optimized. The system will issue a medium-level warning prompt, requiring the team to analyze the emotional characteristic information and record the fluctuations. In addition, a phased analysis report should be generated to track the changes in the strong correlation coefficient, so as to adjust the strategy in a timely manner during future monitoring cycles.

[0189] When a high-risk strong correlation anomaly is detected, it indicates that the model may be severely interfered with, with systemic errors, data anomalies, or extreme fluctuations in emotional states. At this time, the system will issue an urgent red warning, requiring an immediate check and correction of the anomaly source. Strong regularization methods (such as Lasso regression) should be adopted or highly correlated features should be removed to avoid model inaccuracy. At the same time, other data sources can be introduced for cross-validation to ensure the accuracy of the analysis results, and temporary strategies can be implemented based on the monitoring results to ensure the reliability and long-term robustness of the model.

[0190] The present invention effectively solves the problem of multicollinearity caused by strong correlations between multi-dimensional data features by calculating the data consistency offset coefficient and the strong correlation coefficient. This collinearity problem often leads to unstable regression coefficients and weights of the model, affecting the accurate judgment of the independent contribution of emotional features to emotional states. By real-time identifying potential abnormal correlations and combining with set risk thresholds, the system can accurately classify and process emotional feature information when multicollinearity occurs, ensuring the output stability of the model. This mechanism improves the accuracy of emotional state prediction, avoids the volatility of prediction results caused by data correlations, ensures the robustness of the model across different time periods and data sets, and thus improves the accuracy and reliability of emotional monitoring.

[0191] The present invention continuously collects emotional feature information, dynamically generates the data consistency offset coefficient, and establishes an analysis set to support long-term analysis of emotional change trends. By comparing the data in the analysis set with set thresholds, the system can identify the risk levels of emotional feature information and provide personalized response strategies based on this. For example, low-risk anomalies may only require regular emotional management reminders, medium risks will trigger psychological counseling interventions, and high risks will activate an emergency response mechanism. This multi-level risk analysis and response strategy enables the system to continuously and dynamically monitor emotional states, support long-term trend analysis and refined management of emotional features, and ensure that students' emotional states can be managed and regulated in a timely and effective manner.

[0192] Only some exemplary embodiments of the present invention have been described by way of illustration. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. College social negative emotion representation analysis system, characterized in that, It includes a data acquisition and preprocessing module, a feature extraction and anomaly detection module, an intelligent perception and negative emotion representation module, and an anomaly risk assessment and response module: The data acquisition and preprocessing module obtains emotion feature information in real time based on multi-dimensional data sources, cleans and preprocesses the obtained emotion feature information to ensure that different emotion feature information is analyzed on the same dimension; The feature extraction and anomaly detection module extracts key features from the preprocessed emotion feature information, and then performs anomaly analysis on the extracted feature information to detect and mark the atypical associations between emotion feature information; The intelligent perception and negative emotion representation module deploys a pre-trained machine learning model in the emotion monitoring system. Through the intelligent perception ability of the machine learning model, it dynamically perceives the subtle changes in the feature data, identifies potential abnormal fluctuations, and maps them to the negative emotion representation results; The anomaly risk assessment and response module further analyzes the identified potential abnormal fluctuations, judges the risk level of the potential abnormal fluctuations, and formulates different response measures for different potential abnormal fluctuation risk levels; The emotion feature information includes language emotion polarity information and circadian emotion variation information. The language emotion polarity information refers to judging the emotional direction and the degree of emotional intensity by analyzing the vocabulary, grammar, and sentence patterns in language texts or speech expressions. The circadian emotion variation information is used to measure the change range and pattern of an individual's emotional state at different time periods within a day; After obtaining the language emotion polarity information and circadian emotion variation information in the emotion feature information, an anomaly analysis is performed on the language emotion polarity information to generate a language emotion polarity fluctuation index, and an anomaly analysis is performed on the circadian emotion variation information to generate a circadian emotion variation index. The atypical associations between the obtained multi-dimensional data are identified through the generated language emotion polarity fluctuation index and circadian emotion variation index; The language emotion polarity fluctuation index and circadian emotion variation index generated after anomaly analysis obtained through multi-dimensional data sources are input into a pre-trained machine learning model. A data consistency offset coefficient is generated through the machine learning model, and the obtained emotion feature information is intelligently evaluated through the data consistency offset coefficient.

2. The college social negative emotion representation analysis system according to claim 1, characterized in that: The generated data consistency offset coefficient is compared and analyzed with a pre-set data consistency offset coefficient reference threshold to identify the strong correlation between the obtained emotion characteristic information. The analysis results are as follows: If the data consistency offset coefficient is greater than or equal to the pre-set data consistency offset coefficient reference threshold, a potential anomaly correlation is generated, indicating that there is a strong correlation between the obtained emotion characteristic information; If the data consistency offset coefficient is less than the pre-set data consistency offset coefficient reference threshold, an emotion characteristic independence is generated, indicating that there is no strong correlation between the obtained emotion characteristic information.

3. The college social negative emotion characterization analysis system according to claim 2, characterized in that: When there is a strong correlation between the obtained emotion characteristic information and a potential abnormal correlation is generated, continuously obtain the data consistency offset coefficient generated after the abnormal analysis of the emotion characteristic information, establish an analysis set, and compare the data consistency offset coefficient in the analysis set with the first-level reference threshold, the second-level reference threshold, and the data consistency offset coefficient reference threshold. Among them, the second-level reference threshold is greater than the first-level reference threshold, the first-level reference threshold is greater than the data consistency offset coefficient reference threshold. Compare and analyze the data consistency offset coefficient with the second-level reference threshold, the first-level reference threshold, and the data consistency offset coefficient reference threshold. Mark the number of data consistency offset coefficients that are less than the first-level reference threshold and greater than or equal to the data consistency offset coefficient reference threshold as Fa, mark the number of data consistency offset coefficients that are less than the second-level reference threshold and greater than or equal to the first-level reference threshold as Fb, and mark the number of data consistency offset coefficients that are greater than or equal to the second-level reference threshold as Fc; Comprehensively analyze Fa, Fb, and Fc to generate a strong correlation coefficient SCC. The formula is as follows: Wherein, h1, h2, and h3 are respectively preset proportionality coefficients of Fa, Fb, and Fc, and Fa, Fb, and Fc are all greater than 0.

4. The college social negative emotion characterization analysis system according to claim 3, characterized in that: Compare and analyze the generated strong correlation coefficient with the pre-set first strong correlation coefficient reference threshold and the second strong correlation coefficient reference threshold. The analysis results are as follows; If the strong correlation coefficient is less than the first strong correlation coefficient reference threshold, further divide the potential abnormal correlation into a low-risk strong correlation anomaly; If the strong correlation coefficient is greater than or equal to the first strong correlation coefficient reference threshold and less than the second strong correlation coefficient reference threshold, further divide the potential abnormal correlation into a medium-risk strong correlation anomaly: If the strong correlation coefficient is greater than or equal to the second strong correlation coefficient reference threshold, further divide the potential abnormal correlation into a high-risk strong correlation anomaly.

5. The college social negative emotion representation analysis system according to claim 1, wherein The logic for performing abnormal analysis on the language emotion polarity information to generate the language emotion polarity fluctuation index is as follows: Under the detection window, for a given language input, the sentiment polarity score S is extracted through a sentiment analysis model i , and it is converted into time series data. The time series data of the input language text is as follows: T = {t i} = {t1, t2, t3, ……, t n}, where T represents the language text sequence collected within the detection window, t i represents the i-th language content at the time point t, n represents the total number of texts collected within the detection window, and the output sentiment polarity sequence data is as follows: S = {S i} = {S1, S2, ……, S m}, where S represents the sentiment polarity sequence, S i represents the sentiment polarity score of the i-th language content at the time point t, and m represents the total number of language texts within the detection window, that is, how many language text data are analyzed in total, reflecting how many sentiment records are processed within the detection window; Calculate the emotion polarity change rate at each time point on the emotion polarity sequence S, capture the emotion polarity change trend through high-order differences, and enhance the sensitivity to mutations. The calculation expression is as follows: ， where G t represents the gradient of the change in sentiment polarity at time point t, is the second-order difference, which measures the acceleration of the change in sentiment polarity and captures sharp fluctuations, is the first-order difference, which captures the rate of change in sentiment polarity at consecutive time points, and α is a balance coefficient used to adjust the influence of the first-order and second-order gradients; Introduce a non - linear weight function to emphasize the gradient of each emotional polarity change G t for emotional polarity anomalies, and the expression is as follows: In the formula, W(G t ) is the non - linear weight parameter, which gives higher influence to drastic emotional polarity changes and reduces the interference of small fluctuations. β is the weight amplification coefficient used to control the degree of non - linear amplification, and e represents the natural base; Accumulate the non-linear weights at all time points to generate the language emotion polarity fluctuation index. The expression is as follows: In the formula, SPFI represents the language emotion polarity fluctuation index, γ is the non-linear scaling coefficient used to control the influence of the polarity score on the fluctuation index, and max(|S|) is the maximum absolute value in the emotion polarity sequence S.

6. The college social negative emotion characterization analysis system according to claim 1, characterized in that: The logic for performing abnormal analysis on the day-night emotion variation information to generate the day-night emotion variation index is as follows: Under the detection window, continuously collect the emotional data of the target individual, and calibrate the collected emotional data as E t , E t is the emotional state value of the detected individual at time point t; The emotion state fluctuates over time. Therefore, we need to calculate the instantaneous change rate of the time series to capture the amplitude of the emotion change. Using the form of the derivative, define the emotion change rate function. The calculation expression is as follows: where ΔE(t) represents the instantaneous emotional change rate, reflecting the emotional fluctuation amplitude at time point t, and f circ (t) is the circadian rhythm weight function, which is used to consider the periodic influence of day and night. Among them, 2π is a factor used to map the period of one day to the period of trigonometric functions; In the emotion instantaneous change rate ΔE(t), introduce a non-linear fluctuation intensity function to detect abnormal emotion fluctuations, which is used to quantify the abnormal fluctuation amplitude of the emotion. The expression is as follows: In the formula, F anom (t) represents the intensity function of abnormal emotional fluctuations, t0 represents the starting time of emotion monitoring, t1 represents the ending time of emotion monitoring, θ is a non-linear adjustment parameter used to control the sensitivity of emotional fluctuations, λ is an adjustment weight used to control the impact of emotional states on abnormal fluctuations, and g(E(t)) is an emotional state adjustment function, where g(E(t)) = exp(-σE(t)), and in the formula, σ is a parameter for controlling the attenuation rate; Generate the day-night emotion variation index based on the time-weighted result of the emotion abnormal fluctuation intensity function combined with the circadian rhythm. The calculation expression is as follows: In the formula, DEVI represents the diurnal mood variation index, and κ is the circadian rhythm weight adjustment factor.

Images