CN117912710A - Teenager mental health data analysis and early warning system - Google Patents

Abstract

The invention relates to the technical field of mental health analysis, in particular to a teenager mental health data analysis and early warning system which comprises a mood recognition module, a psychological portrait construction module, a risk layering module, a mood prediction module, a behavior analysis module, an intervention strategy module, an early warning feedback module and a health trend monitoring module. In the invention, the clustering analysis and principal component analysis methods are used for analyzing individuals and groups, psychological images are constructed, scientific and reasonable risk classification is realized through a support vector machine and a decision tree classification technology, the prediction accuracy of emotion trend is enhanced by combining an autoregressive integral sliding average model and a circulating neural network, social and learning behavior patterns are deeply understood by means of data mining technology and graph theory analysis, intervention data support is provided, personalized intervention measure formulation is promoted through expert system and rule engine technology, and early warning timeliness and feedback effectiveness are improved through diversified data processing and feedback mechanisms.

Description

Teenager mental health data analysis and early warning system

Technical Field

The invention relates to the technical field of mental health analysis, in particular to a teenager mental health data analysis and early warning system.

Background

The technical field of mental health analysis is focused on understanding and predicting mental health of individuals using data analysis methods and tools. Within this field, data science and psychological principles are combined to parse mental health data collected from various sources, including behavioral patterns, social interactions, feelings of self-results, and emotions, etc., with the aim of providing insight into the mental state of individuals.

The teenager mental health data analysis and early warning system is a tool integrating data analysis technology and psychological principles and aims at monitoring and evaluating the mental health state of the teenager. The main purpose of this system is to discover mental health problems faced by teenagers, such as depression, anxiety problems, in time by analyzing relevant data, such as mood changes, social behavior, school achievements, etc. By early identification of these problems, the system aims to facilitate timely intervention and support, thereby preventing exacerbations of the problem.

The traditional mental health system for teenagers has been shown to be inadequate in many respects. The construction of psychological images lacks an efficient data processing method, and the psychological characteristics of individuals and groups are difficult to accurately reflect. In terms of risk assessment, traditional systems lack a refined, personalized risk classification, affecting the pertinence of preventive and interventional measures. The emotion prediction is based on a simple statistical method, and lacks advanced algorithm support, so that the reliability of a prediction result is limited. Behavior analysis ignores deep mining of social media and learning behavior data, resulting in incomplete behavior pattern analysis. The formulation of intervention strategies lacks personalized and systematic designs, and effective intervention is difficult to achieve.

Disclosure of Invention

The invention aims to solve the defects in the prior art and provides a teenager mental health data analysis and early warning system.

In order to achieve the above purpose, the present invention adopts the following technical scheme: the teenager mental health data analysis and early warning system comprises an emotion recognition module, a mental portrait construction module, a risk layering module, an emotion prediction module, a behavior analysis module, an intervention strategy module, an early warning feedback module and a health trend monitoring module;

the emotion recognition module is used for carrying out word boundary recognition and emotion tendency analysis by adopting a natural language processing technology based on text, voice and facial expression data, carrying out facial expression feature extraction and voice emotion feature recognition by adopting a deep learning technology, integrating multi-modal information by an integrated model, and generating an emotion recognition result by parameter tuning and cross verification accuracy after model training;

The psychological portrait construction module is used for carrying out homogeneous group recognition by adopting cluster analysis based on emotion recognition results, carrying out feature dimension reduction by adopting principal component analysis, synthesizing individual historical data and behavior features in portrait construction, and generating a psychological portrait by selecting and matching differential group features through an optimization algorithm;

The risk layering module is used for carrying out individual risk division by adopting a support vector machine based on psychological representation, adopting decision tree group risk classification, and then adopting grid search optimization super-parameters in a classification model to generate a risk grade result;

The emotion prediction module processes time series data through an autoregressive moving average model based on historical data and risk grade results, carries out trend and seasonal decomposition, and captures long-term dependency relations through a circulating neural network to generate emotion trend prediction;

The behavior analysis module performs pattern discovery and association rule learning based on social media and learning behavior data by a data mining technology, and graph theory analysis reveals a relationship network structure and generates a behavior analysis result;

The intervention strategy module provides decision support based on a behavior analysis result and emotion trend prediction and an expert system, the rule engine executes a strategy based on set conditions, and the strategy generation process refers to personalized requirements and real-time data updating to generate an intervention strategy;

the early warning feedback module integrates feedback information based on a risk level result and an intervention strategy by adopting multivariate data processing, performs effect evaluation and user response based on a feedback mechanism, and generates early warning notification and feedback by adopting a data-driven decision-making cycle to ensure updating timeliness and early warning accuracy;

the health trend monitoring module is used for identifying abnormal and predicted risks through pattern recognition based on long-term data and trend analysis to reveal a change pattern, carrying out long-term monitoring and model updating, and generating a health trend result based on comprehensive evaluation of health indexes.

As a further scheme of the invention, the emotion recognition result is specifically emotion state distribution, emotion intensity analysis and emotion type recognition, the psychological portrait comprises individual behavior characteristics, group behavior trends and psychological state dynamics, the risk level result is specifically high risk, medium risk and low risk classification, the emotion trend prediction comprises short-term emotion change, long-term emotion trend and potential emotion risk, the behavior analysis result is specifically social interaction mode, learning habit analysis and daily activity preference, the early warning notification and feedback is specifically early warning message distribution, user feedback collection and early warning effect analysis, and the health trend result is specifically long-term psychological health change, key health index and potential risk prediction.

As a further scheme of the invention, the emotion recognition module comprises a text emotion analysis sub-module, a voice emotion analysis sub-module and a facial expression analysis sub-module;

The text emotion analysis submodule is used for carrying out text emotion analysis by adopting a natural language processing technology based on text data, receiving user text through a text crawler or direct input in a data input stage, carrying out part-of-speech tagging, syntactic analysis and emotion vocabulary library matching in a processing stage, carrying out emotion classification and strength evaluation by using a machine learning algorithm, and generating a text emotion analysis result;

The voice emotion analysis submodule adopts a deep learning technology to carry out emotion analysis on voice data, the voice data is input through a microphone or a recording file, feature extraction comprises voice frequency, rhythm and tone analysis, and a convolutional neural network and a cyclic neural network are used for carrying out emotion state recognition to generate a voice emotion analysis result;

The facial expression analysis submodule analyzes facial expression data by adopting an image processing technology, the facial expression data is input through a camera or an image file, the analysis process comprises facial feature point recognition and expression dynamic analysis, and emotion judgment is carried out by utilizing a statistical model and a machine learning method to generate a facial expression emotion analysis result.

As a further scheme of the invention, the psychological portrait construction module comprises an individual portrait sub-module, a group portrait sub-module and a behavioral pattern analysis sub-module;

The individual portrait sub-module analyzes the emotion and behavior modes of the individual by adopting a clustering analysis method based on the emotion recognition result, specifically adopts a K-means algorithm and a hierarchical clustering method, improves the clustering effect by iteratively optimizing a clustering center and a hierarchical structure, determines the optimal clustering quantity based on a contour coefficient index, and generates an individual psychological portrait;

The group portrait submodule adopts a principal component analysis method to conduct group feature analysis based on the individual psychological portraits, specifically conducts linear dimension reduction to extract principal variables and key features in group data, selects principal component number maximization information retention, evaluates the dimension selection effect through accumulated contribution rate, and generates the group psychological portraits;

The behavior pattern analysis submodule is used for carrying out deep individual and group behavior analysis through a behavior pattern recognition technology based on the group psychological portrait, the analysis comprises time sequence analysis and association rule mining, the behavior trend is predicted based on the time sequence analysis, the association rule mining discovers potential relations among behaviors, and a behavior pattern analysis result is generated.

As a further scheme of the invention, the risk layering module comprises a risk factor identification sub-module, a risk grade classification sub-module and a risk early warning sub-module;

The risk factor identification sub-module is based on psychological representation, adopts a support vector machine algorithm, performs kernel function conversion to perform high-dimensional mapping, then performs maximum interval classification to identify and analyze risk factors, and simultaneously performs individual behavior and psychological characteristic analysis by using a data mining technology to generate a risk factor identification result;

The risk level classification submodule adopts a decision tree classification technology based on the risk factor identification result, generates and branches and clips by using classification rules, and combines information gain and a keni unrepeace index to carry out risk level classification to generate a risk level classification result;

the risk early warning sub-module is used for carrying out risk monitoring by adopting a correlation analysis method based on a risk grade classification result, collecting and arranging time sequence data, carrying out correlation analysis by adopting a pearson correlation coefficient and a spearman grade correlation, and continuously monitoring risk changes by combining trend analysis and anomaly detection technology to generate a risk early warning result.

As a further scheme of the invention, the emotion prediction module comprises a time sequence analysis sub-module, a trend prediction sub-module and a prediction result analysis sub-module;

The time sequence analysis submodule performs data import based on historical data, performs emotion change trend analysis by adopting an autoregressive integral moving average model, comprises difference, parameter estimation and model diagnosis, and combines seasonal adjustment and outlier correction to generate a time sequence analysis result;

The trend prediction submodule predicts the emotion trend in the future based on the time sequence analysis result and the risk level result by adopting a circulating neural network and based on a long-short-period memory network and a gating circulating unit, and comprises network structure definition, parameter initialization and iterative training, predicts by utilizing history and current data, and generates an emotion trend prediction result;

The prediction result analysis submodule is used for carrying out data analysis by using a data analysis and pattern recognition technology including a statistical analysis and data visualization method based on the emotion trend prediction result, adopting cluster analysis and association rule mining to identify and refine key emotion driving factors, and carrying out evaluation and optimization to refine analysis results so as to generate a prediction analysis result.

As a further scheme of the invention, the behavior analysis module comprises a social behavior analysis sub-module, a learning behavior analysis sub-module and a daily behavior analysis sub-module;

The social behavior analysis submodule adopts K-means clustering to conduct feature vectorization on users based on social media data, iteratively optimizes a cluster center to form a user group, then applies a frequent pattern mining algorithm to conduct item set generation and support calculation, identifies a key social behavior pattern and generates a social behavior analysis result;

the learning behavior analysis sub-module analyzes the learning behavior sequence by adopting sequential pattern mining based on the social behavior analysis result, discovers a repeated pattern or trend, determines key learning behavior characteristics by combining decision tree analysis, performs attribute selection and tree construction, classifies and understands differentiated learning patterns, and generates a learning behavior analysis result;

The daily behavior analysis submodule measures the position and the action of an individual in a social network by adopting network centrality analysis based on daily life data, performs shortest path calculation and path dependence analysis by using a path analysis method, reveals the rule and the mode of the daily behavior of the individual, and generates a daily behavior analysis result.

As a further scheme of the invention, the intervention strategy module comprises an individuation intervention sub-module, a group intervention strategy sub-module and an intervention effect evaluation sub-module;

The individualized intervention sub-module builds a rule base and sets an inference mechanism based on a behavior analysis result and an emotion trend prediction result by adopting a rule-based expert system, and performs emotion and behavior pattern analysis by combining with fuzzy logic, including definition of a fuzzy set and evaluation of fuzzy rules, so as to generate an individualized intervention strategy;

The group intervention strategy sub-module integrates prediction results of a plurality of models by adopting a random forest and gradient lifting mechanism based on an individuation intervention strategy, simultaneously applies social network analysis to analyze network structures and roles, identifies key social dynamics and group behavior modes and generates a group intervention strategy;

the intervention effect evaluation submodule carries out regression analysis of intervention effect based on a group intervention strategy, comprises linear and nonlinear models, matches multiple types of data characteristics and complexity of the intervention effect, carries out comprehensive evaluation by combining objective scores and subjective feedback, analyzes influence of intervention, and generates intervention effect evaluation.

As a further scheme of the invention, the early warning feedback module comprises an early warning notification sub-module, a feedback collection sub-module and an early warning effect analysis sub-module;

The early warning notification sub-module uses a data-driven notification mechanism based on a risk level result and an intervention strategy, comprises customizing message content and a user behavior tracking algorithm, adopts a behavior pattern recognition technology, recognizes a key behavior pattern, optimizes an information sending strategy, automatically sends early warning information and tracks user response, and generates an early warning notification result;

The feedback collection sub-module collects user feedback through a response type design questionnaire based on an early warning notification result by using an online investigation tool and a feedback analysis algorithm, analyzes the collected data by adopting a text mining and emotion analysis technology, processes the user feedback in real time, and generates a feedback collection result;

The early warning effect analysis submodule adopts a statistical analysis method and an effect evaluation model based on feedback collection results, quantitatively analyzes the collected data through a multivariate analysis and hypothesis testing statistical method, evaluates the effect of intervention measures, comprehensively evaluates the intervention effect by using an effect scoring algorithm and a user satisfaction evaluation model, and generates early warning effect analysis.

As a further scheme of the invention, the health trend monitoring module comprises a long-term trend analysis sub-module, a health index monitoring sub-module and a risk prediction sub-module;

The long-term trend analysis submodule analyzes historical trend of mental health based on long-term data by adopting a time sequence analysis technology and an adaptive prediction algorithm and a linear regression model, predicts future changes and generates a long-term trend analysis result;

The health index monitoring submodule is used for carrying out real-time monitoring on health indexes by combining abnormal behavior detection and daily activity analysis on the basis of long-term trend analysis results and adopting physiological signal monitoring technologies including heart rate monitoring and sleep quality analysis, and generating health index monitoring results through continuous monitoring and behavior data analysis;

the risk prediction submodule performs health risk prediction and assessment by applying a machine learning technology and a risk assessment model, wherein the machine learning technology and the risk assessment model comprise nonlinear regression analysis and neural network prediction, probability risk scoring and influence factor weight analysis, and generates a risk prediction result.

Compared with the prior art, the invention has the advantages and positive effects that:

In the invention, the clustering analysis and principal component analysis methods are used for carrying out deep analysis on individuals and groups to construct more accurate psychological images. Scientific and reasonable risk classification is realized through a support vector machine and a decision tree classification technology. And combining an autoregressive integral moving average model and a circulating neural network, so that the prediction accuracy of the emotion trend is enhanced. Based on data mining technology and graph theory analysis, social and learning behavior patterns are deeply understood, and data support is provided for intervention. Personalized and efficient intervention measure formulation is promoted through expert system and rule engine technology. Through diversified data processing and feedback mechanisms, timeliness of early warning and effectiveness of feedback are improved.

Drawings

FIG. 1 is a system flow diagram of the present invention;

FIG. 2 is a schematic diagram of a system framework of the present invention;

FIG. 3 is a flowchart of an emotion recognition module of the present invention;

FIG. 4 is a flow chart of a psychological image construction module of the present invention;

FIG. 5 is a flow chart of a risk stratification module of the present invention;

FIG. 6 is a flowchart of an emotion prediction module of the present invention;

FIG. 7 is a flow chart of a behavior analysis module according to the present invention;

FIG. 8 is a flow chart of an intervention strategy module of the present invention;

FIG. 9 is a flow chart of the early warning feedback module of the present invention;

FIG. 10 is a flow chart of a health trend monitoring module according to the present invention.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

In the description of the present invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like indicate orientations or positional relationships based on the orientation or positional relationships shown in the drawings, merely to facilitate describing the present invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, in the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.

Embodiment one: referring to fig. 1 to 2, the teenager mental health data analysis and early warning system comprises a mood recognition module, a mental portrait construction module, a risk layering module, a mood prediction module, a behavior analysis module, an intervention strategy module, an early warning feedback module and a health trend monitoring module;

The emotion recognition module is used for carrying out word boundary recognition and emotion tendency analysis by adopting a natural language processing technology based on text, voice and facial expression data, carrying out facial expression feature extraction and voice emotion feature recognition by adopting a deep learning technology, integrating multi-modal information by an integrated model, and generating an emotion recognition result by parameter tuning and cross verification accuracy after model training;

the psychological portrait construction module carries out homogeneous group recognition by adopting cluster analysis based on emotion recognition results, carries out feature dimension reduction by adopting principal component analysis, synthesizes individual history data and behavior features in portrait construction, and selects and matches differential group features by an optimization algorithm to generate a psychological portrait;

The risk layering module is used for carrying out individual risk division by adopting a support vector machine based on the psychological portrait, adopting decision tree group risk classification, and then adopting grid search optimization super-parameters in a classification model to generate a risk grade result;

the emotion prediction module processes time series data through an autoregressive moving average model based on historical data and risk grade results, carries out trend and seasonal decomposition, and captures long-term dependence by a circulating neural network to generate emotion trend prediction;

The behavior analysis module performs pattern discovery and association rule learning based on social media and learning behavior data by a data mining technology, and graph theory analysis reveals a relationship network structure and generates a behavior analysis result;

The intervention strategy module provides decision support based on a behavior analysis result and emotion trend prediction and an expert system, the rule engine executes a strategy based on set conditions, and the strategy generation process refers to personalized requirements and real-time data updating to generate an intervention strategy;

The early warning feedback module integrates feedback information based on a risk level result and an intervention strategy by adopting multi-element data processing, performs effect evaluation and user response based on a feedback mechanism, and generates early warning notification and feedback by adopting a data-driven decision-making cycle to ensure updating timeliness and early warning accuracy;

The health trend monitoring module is used for identifying abnormal and predicted risks through pattern identification based on long-term data and trend analysis to reveal a change pattern, carrying out long-term monitoring and model updating, and generating a health trend result based on comprehensive evaluation of health indexes.

The emotion recognition results comprise emotion state distribution, emotion intensity analysis and emotion type recognition, the psychological portrait comprises individual behavior characteristics, group behavior trends and psychological state dynamics, the risk level results comprise high risk, medium risk and low risk classification, the emotion trend prediction comprises short-term emotion change, long-term emotion trend and potential emotion risk, the behavior analysis results comprise social interaction modes, learning habit analysis and daily activity preference, the early warning notification and feedback comprise early warning message distribution, user feedback collection and early warning effect analysis, and the health trend results comprise long-term psychological health change, key health indexes and potential risk prediction.

Through the accurate emotion state analysis of the emotion recognition module, the individual and group behavior characteristic insight of the psychological portrait construction module and the detailed risk assessment of the risk layering module, the system provides personalized support and intervention for teenagers. The emotion prediction module makes future emotion changes and potential risks predictable, so that effective measures are taken before the problem is exacerbated. Deep analysis of the behavior analysis module reveals social and learning modes of teenagers and provides key information for understanding psychological health influence factors of the teenagers. The intervention strategy module combines an expert system and a rule engine, provides an accurate and personalized intervention strategy, and enhances timeliness and pertinence of intervention. The early warning feedback module ensures the accuracy and timeliness of the early warning system through a real-time feedback mechanism. The long-term monitoring and model updating of the health trend monitoring module are beneficial to continuously tracking the mental health state of teenagers and preventing long-term or new mental health risks.

Referring to fig. 3, the emotion recognition module includes a text emotion analysis sub-module, a voice emotion analysis sub-module, and a facial expression analysis sub-module;

The text emotion analysis submodule is used for carrying out text emotion analysis by adopting a natural language processing technology based on text data, receiving user text through a text crawler or direct input in a data input stage, carrying out part-of-speech tagging, syntax analysis and emotion vocabulary library matching in a processing stage, carrying out emotion classification and strength evaluation by using a machine learning algorithm, and generating a text emotion analysis result;

The voice emotion analysis submodule adopts a deep learning technology to carry out emotion analysis on voice data, the voice data is input through a microphone or a recording file, feature extraction comprises voice frequency, rhythm and tone analysis, and a convolutional neural network and a cyclic neural network are used for carrying out emotion state recognition to generate a voice emotion analysis result;

The facial expression analysis submodule analyzes facial expression data by adopting an image processing technology, the facial expression data is input through a camera or an image file, the analysis process comprises facial feature point recognition and expression dynamic analysis, and emotion judgment is carried out by utilizing a statistical model and a machine learning method to generate a facial expression emotion analysis result.

In the text emotion analysis sub-module, natural language processing techniques perform emotion analysis based on user input or raw text data obtained through a text crawler. The process encompasses part-of-speech tagging, identifying various grammatical elements in text, such as nouns, verbs, and the like. Then, syntactic analysis is performed to determine the association between words and sentence structure. The emotion vocabulary library matching technique is then initiated to evaluate the emotion expressions in the text. The machine learning algorithm, such as a support vector machine or a neural network, carries out emotion classification and strength evaluation on the text, judges the text to be positive, negative or neutral, and the text emotion analysis result generated in the process is extremely important for understanding the emotion state of the user, especially in the fields of customer service, market analysis and the like.

In the speech emotion analysis sub-module, deep learning techniques are used to perform emotion analysis on real-time speech captured by a microphone or pre-recorded audio files. The feature extraction links focus on the frequency, rhythm, and pitch of the sound. Deep learning algorithms such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) then analyze these features to identify emotional states in the speech. The generated voice emotion analysis result reflects the emotion state of the speaker and has wide application in the fields of real-time communication, emotion calculation and the like.

In the facial expression analysis sub-module, image processing techniques analyze real-time images captured by a camera or stored image files. The process begins with the identification of facial feature points, such as the locations of eyes, mouth, and nose, followed by expression dynamics analysis. Statistical models and machine learning methods, such as decision trees or random forest algorithms, are used to make emotion decisions on facial expressions. The generated facial expression emotion analysis result has important value for enhancing the man-machine interaction and emotion recognition system.

Referring to fig. 4, the psychological portrait construction module includes an individual portrait sub-module, a group portrait sub-module, and a behavioral pattern analysis sub-module;

the individual portrait sub-module analyzes the emotion and behavior modes of the individual by adopting a clustering analysis method based on the emotion recognition result, specifically adopts a K-means algorithm and a hierarchical clustering method, improves the clustering effect by iteratively optimizing a clustering center and a hierarchical structure, determines the optimal clustering quantity based on the contour coefficient index, and generates an individual psychological portrait;

The group image sub-module is used for carrying out group feature analysis by adopting a principal component analysis method based on the individual psychological images, specifically carrying out linear dimension reduction to extract principal variables and key features in group data, selecting principal component number maximizing information retention, and evaluating the dimension selection effect by accumulating contribution rates to generate group psychological images;

The behavior pattern analysis submodule performs deep individual and group behavior analysis through a behavior pattern recognition technology based on the group psychological portrait, the analysis comprises time sequence analysis and association rule mining, the behavior trend is predicted based on the time sequence analysis, and the association rule mining discovers potential relations among behaviors and generates a behavior pattern analysis result.

In the individual representation sub-module, the data format is based primarily on the results of emotion recognition, which typically includes data about the individual's emotional state and behavioral patterns, such as emotional intensity, frequency, etc. The submodule adopts a clustering analysis method, in particular a K-mean algorithm and a hierarchical clustering method. The K-means algorithm classifies data into a plurality of categories by iteratively optimizing a cluster center, while the hierarchical clustering algorithm organizes data by constructing a hierarchical structure. In the operation process, the emotion and behavior modes are initially clustered through a K-means algorithm, and then the clusters are further refined and optimized through a hierarchical clustering method. In determining the optimal number of clusters, a profile factor index is employed, which helps to evaluate the quality of the clusters. The result of this series of operations is the generation of detailed psychological representations of the individual that reflect the emotional and behavioral characteristics of the individual, providing important information for understanding the individual's behavioral patterns.

In the group image sub-module, the processed data is based on the individual psychological image generated in the previous step. The module uses Principal Component Analysis (PCA) to process the data. PCA is mainly a linear dimension reduction technique, and simplifies the data set by extracting main variables and key features in the data. In the execution process, the individual psychological image data is subjected to linear dimension reduction, and key characteristics and main variables are extracted. The appropriate number of principal components is selected to maximize the retention of information. The validity of the selected dimension is assessed by calculating the cumulative contribution. Through these steps, group psychological images are generated which reveal the core psychological characteristics and behavioral tendencies of the group, which are critical to understanding and predicting group behavioral patterns.

In the behavior pattern analysis sub-module, the input data is a mental representation of the population. The module adopts behavior pattern recognition technology, including time sequence analysis and association rule mining. Time series analysis is used to predict behavioral trends, and future behavioral patterns are predicted by analyzing the change of data over time. Association rule mining is then used to discover potential relationships between behaviors, revealing potential behavior patterns by analyzing relationships between different behaviors. These analytical methods help to understand the behavioral characteristics of individuals and populations in depth. This sub-module generates behavioral pattern analysis results that provide powerful support for understanding and predicting individual and group behaviors.

Referring to fig. 5, the risk stratification module includes a risk factor identification sub-module, a risk level classification sub-module, and a risk early warning sub-module;

The risk factor identification sub-module is based on psychological representation, adopts a support vector machine algorithm, performs kernel function conversion to perform high-dimensional mapping, then performs maximum interval classification to identify and analyze risk factors, and simultaneously performs individual behavior and psychological characteristic analysis by using a data mining technology to generate a risk factor identification result;

The risk level classification sub-module adopts a decision tree classification technology based on the risk factor identification result, generates and branches and clips by using a classification rule, and combines information gain and a keni unrepeace index to carry out risk level classification to generate a risk level classification result;

the risk early warning sub-module is used for carrying out risk monitoring by adopting a correlation analysis method based on a risk level classification result, collecting and arranging time sequence data, carrying out correlation analysis by adopting a pearson correlation coefficient and a spearman level correlation, and continuously monitoring risk changes by combining trend analysis and anomaly detection technology to generate a risk early warning result.

In the risk factor identification sub-module, the risk factors are identified by using a Support Vector Machine (SVM) algorithm through the psychological portrait data. This process involves kernel function conversion to achieve mapping of high-dimensional data in order to more effectively separate the data in a higher-dimensional space. In performing maximum interval classification, the SVM algorithm finds decision boundaries that maximize the interval between different classes. In addition, the submodule further utilizes a data mining technology to conduct deep analysis on the behavioral and psychological characteristics of the individual. This series of operations makes it possible to identify key factors related to risk and generate a risk factor identification result. These results are critical for understanding the risk characteristics of individuals and for taking precautions in advance.

In the risk level classification sub-module, a decision tree classification technology is adopted to divide risk levels based on the risk factor identification result. By applying the method of classifying rules to generate and branch clips, risk factors are effectively classified into different grades based on information gain and a genie unrepeatation index. This classification approach makes the judgment of risks more explicit and orderly, helping to better understand and manage these risks. The generated risk level classification result provides an important decision basis for subsequent risk management and strategy countermeasures.

And in the risk early warning sub-module, based on the risk grade classification result, performing risk monitoring by adopting a correlation analysis method. Time series data are collected and collated, and correlation analysis is performed on the data by applying a pearson correlation coefficient and a spearman rank correlation analysis method. In addition, by combining trend analysis and anomaly detection technology, potential risks are effectively identified, and the process not only monitors the current risk condition, but also predicts the risk trend appearing in the future. The generated risk early warning result is real-time reflection of the continuously-changing risk condition, and key information is provided for timely response and preventive measures.

Referring to fig. 6, the emotion prediction module includes a time sequence analysis sub-module, a trend prediction sub-module, and a prediction result analysis sub-module;

The time sequence analysis submodule performs data import based on historical data, performs emotion change trend analysis by adopting an autoregressive integral moving average model, comprises difference, parameter estimation and model diagnosis, and combines seasonal adjustment and outlier correction to generate a time sequence analysis result;

The trend prediction sub-module predicts the future emotion trend based on the time sequence analysis result and the risk level result by adopting a circulating neural network and based on a long-short-period memory network and a gating circulating unit, and comprises network structure definition, parameter initialization and iterative training, and predicts by utilizing historical and current data to generate an emotion trend prediction result;

The prediction result analysis submodule is used for carrying out data analysis by using a data analysis and pattern recognition technology including a statistical analysis and data visualization method based on the emotion trend prediction result, adopting cluster analysis and association rule mining to identify and refine key emotion driving factors, and carrying out evaluation and optimization to refine the analysis result so as to generate a prediction analysis result.

In the time series analysis sub-module, an autoregressive integrated moving average model (ARIMA) is employed.

ARIMA model

Example code (Python):

import pandas as pdfrom statsmodels.tsa.arima.model import ARIMA

# data import

data = pd.read_csv('emotional_data.csv')

The #ARIMA model def apply_arima (series):

model = ARIMA(series, order=(5, 1, 0))

model_fit = model.fit()

return model_fit

series = data['emotion_metric']

arima_result = apply_arima(series)

In the trend prediction submodule, a Recurrent Neural Network (RNN) uses a long short term memory network (LSTM) and a gated loop unit (GRU).

LSTM network

Example code (Python):

from keras.models import Sequentialfrom keras.layers import LSTM, Dense

def build_lstm_model(input_shape):

model = Sequential()

model.add(LSTM(50, activation='relu', input_shape=input_shape))

model.add(Dense(1))

model.compile(optimizer='adam', loss='mse')

return model

# assume that X_train and y_train are processed training data

lstm_model = build_lstm_model((X_train.shape[1], 1))

lstm_model.fit(X_train, y_train, epochs=100, batch_size=32)

In the prediction result analysis sub-module, data analysis and pattern recognition (cluster analysis and association rule mining) are performed.

Cluster analysis

Example code (Python):

from sklearn.cluster import KMeansimport matplotlib.pyplot as pltimport seaborn as sns

def apply_kmeans(data):

kmeans = KMeans(n_clusters=3)

kmeans.fit(data)

return kmeans.labels_

labels = apply_kmeans(data)

# data visualization

plt.figure(figsize=(10, 6))

sns.scatterplot(data=data, x='feature_1', y='feature_2', hue=labels)

plt.title('Emotion Clustering Results')

plt.show()

Referring to fig. 7, the behavior analysis module includes a social behavior analysis sub-module, a learning behavior analysis sub-module, and a daily behavior analysis sub-module;

The social behavior analysis submodule adopts K-means clustering to conduct feature vectorization on users based on social media data, iteratively optimizes a cluster center to form a user group, then applies a frequent pattern mining algorithm to conduct item set generation and support calculation, identifies a key social behavior pattern and generates a social behavior analysis result;

The learning behavior analysis sub-module analyzes the learning behavior sequence based on the social behavior analysis result by adopting sequential pattern mining, discovers a repeated pattern or trend, determines key learning behavior characteristics by combining decision tree analysis, performs attribute selection and tree construction, classifies and understands differentiated learning patterns, and generates a learning behavior analysis result;

the daily behavior analysis submodule measures the position and the action of an individual in a social network by adopting network centrality analysis based on daily life data, performs shortest path calculation and path dependence analysis by using a path analysis method, reveals the rule and the mode of the daily behavior of the individual, and generates a daily behavior analysis result.

And in the social behavior analysis submodule, carrying out feature vectorization on the user through social media data, and adopting a K-mean clustering algorithm. The algorithm forms different user groups by iteratively optimizing the cluster center. In this process, the social behavior data of the user is converted into feature vectors, which are then grouped by a K-means clustering algorithm to form different user groups. And then applying a frequent pattern mining algorithm to carry out deep analysis on the user group, including item set generation and support calculation, so as to identify a key social behavior pattern. The result of this operation is the generation of social behavior analysis results that reveal typical patterns of behavior of the user on social media, which are of great significance for understanding social media user behavior and trends.

In the learning behavior analysis sub-module, a sequence of learning behaviors is analyzed using a sequential pattern mining method based on the results of the social behavior analysis. This approach enables finding repetitive patterns or trends in the learning process. On this basis, key features of learning behavior, including attribute selection and tree construction, are determined in conjunction with decision tree analysis. Such processing not only classifies different learning patterns, but also facilitates a deep understanding of the features of these patterns. The generated learning behavior analysis result provides an important view for understanding the behavior pattern in the learning process, and is helpful for designing more effective teaching strategies and learning plans.

In the daily behavior analysis sub-module, the position and the effect of an individual in a social network are measured by analyzing daily life data and adopting network centrality analysis. This analysis includes shortest path computation and path-dependent analysis, aimed at revealing the rules and patterns of the individual's daily behavior. Network centrality analysis highlights the importance and impact of individuals in a social network, while path analysis reveals interrelationships and dependencies between behaviors. The generated daily behavior analysis result deeply reveals the behavior characteristics and social patterns of the individual in daily life, and precious information is provided for understanding the social behavior and daily habit of the individual.

Referring to fig. 8, the intervention policy module includes a personalized intervention sub-module, a group intervention policy sub-module, and an intervention effect evaluation sub-module;

based on the behavior analysis result and the emotion trend prediction result, the individuation intervention sub-module adopts a rule-based expert system to construct a rule base, sets an inference mechanism, combines with fuzzy logic to analyze emotion and behavior patterns, and comprises definition of fuzzy sets and evaluation of fuzzy rules to generate individuation intervention strategies;

The group intervention strategy sub-module integrates prediction results of a plurality of models by adopting a random forest and gradient lifting mechanism based on an individuation intervention strategy, simultaneously applies social network analysis to analyze network structures and roles, identifies key social dynamics and group behavior modes and generates a group intervention strategy;

The intervention effect evaluation submodule carries out regression analysis of intervention effect based on a group intervention strategy, wherein the regression analysis comprises linear and nonlinear models, the complexity of matching multiple types of data characteristics and the intervention effect is comprehensively evaluated by combining objective scores and subjective feedback, and the influence of intervention is analyzed to generate intervention effect evaluation.

In the personalized intervention sub-module, a rule-based expert system is adopted to construct a personalized intervention strategy through a behavior analysis result and an emotion trend prediction result. An enriched rule base is built in the expert system, and a corresponding reasoning mechanism is set. These rule bases contain various analyses of emotion and behavior patterns, processing uncertainties and ambiguities based on fuzzy logic. In a specific operation, fuzzy sets are defined, then fuzzy rules are evaluated and applied, the process involves converting behavioral and emotional data into fuzzy values, and then analysis is performed through a rule base in an expert system to generate an individualized intervention strategy. These strategies can provide customized interventions for an individual's specific needs, helping to more effectively manage and improve the individual's mood and behavior.

In the population intervention strategy sub-module, based on the personalized intervention strategy, the prediction results of the multiple models are integrated using a random forest and gradient lifting mechanism. These ensemble learning methods can improve the accuracy and robustness of predictions. The social network analysis is applied to understand the network structure and roles, and the key social dynamics and group behavior patterns are identified, in the process, not only the behavior characteristics of individuals in the group are considered, but also the social connection and influence distribution inside the group are analyzed. The generated group intervention strategy provides an effective means for managing and optimizing group behaviors, and is helpful for promoting positive interaction and behavior improvement inside the group.

In the intervention effect evaluation sub-module, regression analysis of intervention effects is performed based on a population intervention strategy. This includes the application of linear and nonlinear models to match the complexity of the multiple classes of data characteristics and intervention effects. In operation, comprehensive evaluation is performed by comparing the data changes before and after intervention and combining objective scores and subjective feedback. The comprehensive evaluation method can analyze the influence of the intervention from multiple angles, and the generated intervention effect evaluation result provides important basis for evaluating and improving the intervention strategy. Through the evaluation results, the effect of the intervention measures can be more accurately understood, so that future intervention strategy optimization and adjustment are guided.

Referring to fig. 9, the early warning feedback module includes an early warning notification sub-module, a feedback collection sub-module, and an early warning effect analysis sub-module;

The early warning notification sub-module uses a data-driven notification mechanism based on the risk level result and the intervention strategy, comprises customizing message content and a user behavior tracking algorithm, adopts a behavior pattern recognition technology, recognizes a key behavior pattern, optimizes an information sending strategy, automatically sends early warning information and tracks user response, and generates an early warning notification result;

The feedback collection sub-module collects user feedback through a response type design questionnaire by using an online investigation tool and a feedback analysis algorithm based on the early warning notification result, analyzes the collected data by adopting a text mining and emotion analysis technology, processes the user feedback in real time, and generates a feedback collection result;

Based on feedback collection results, the early warning effect analysis submodule adopts a statistical analysis method and an effect evaluation model, quantitatively analyzes the collected data through a multivariate analysis and hypothesis testing statistical method, evaluates the effect of intervention measures, comprehensively evaluates the intervention effect by using an effect scoring algorithm and a user satisfaction evaluation model, and generates early warning effect analysis.

And in the early warning notification sub-module, automatically sending early warning information by using a data-driven notification mechanism based on the risk level result and the intervention strategy. The mechanism includes customized message content and user behavior tracking algorithms, as well as behavior pattern recognition techniques. In actual operation, key behavior modes are identified, and information sending strategies are optimized according to the modes, so that the relevance and timeliness of information are ensured. Automatically sending early warning information and tracking the reaction of the user. This approach enables efficient communication of important information and real-time monitoring of the user's response. The generated early warning notification result is helpful for evaluating the validity of the notification strategy and provides basis for future optimization.

In the feedback collection sub-module, user feedback is collected using an online survey tool and a feedback analysis algorithm based on the early warning notification results. Feedback information of the user is collected through a responsive designed questionnaire. The collected data is subjected to deep analysis by adopting text mining and emotion analysis technology, so that the direct feedback of a user can be understood, and emotion and attitude can be obtained. The process of processing user feedback in real time enables quick response to user demands and questions, and the generated feedback collection results are critical for understanding user satisfaction and acceptance of early warning notices.

And in the early warning effect analysis submodule, based on feedback collection results, adopting a statistical analysis method and an effect evaluation model to carry out deep analysis. The collected data is quantitatively analyzed by multivariate analysis and hypothesis testing statistical methods to evaluate the success of the intervention. And comprehensively evaluating by using an effect scoring algorithm and a user satisfaction evaluation model, wherein the comprehensive evaluation can comprehensively reflect the effect of the intervention measures. The generated early warning effect analysis result provides a quantitative basis for evaluating and improving the intervention measures, so that the intervention strategy can respond to the demands of users more accurately and effectively.

Referring to fig. 10, the health trend monitoring module includes a long-term trend analysis sub-module, a health index monitoring sub-module, and a risk prediction sub-module;

The long-term trend analysis submodule analyzes historical trend of mental health by adopting a self-adaptive prediction algorithm and a linear regression model based on long-term data and adopting a time sequence analysis technology, predicts future change and generates a long-term trend analysis result;

The health index monitoring submodule is used for carrying out real-time monitoring on health indexes by adopting physiological signal monitoring technologies including heart rate monitoring and sleep quality analysis and combining abnormal behavior detection and daily activity analysis based on long-term trend analysis results, and generating health index monitoring results through continuous monitoring and behavior data analysis;

The risk prediction sub-module performs health risk prediction and assessment by applying a machine learning technology and a risk assessment model, including nonlinear regression analysis and neural network prediction, probability risk scoring and influence factor weight analysis, based on the health index monitoring result, so as to generate a risk prediction result.

In the long-term trend analysis sub-module, a time series analysis technology is adopted to explore the historical trend of mental health by analyzing long-term data, and the process relates to the application of an adaptive prediction algorithm and a linear regression model. Time series analysis techniques process long-term collected mental health data to identify and understand past trends and patterns. And predicting future changes through an adaptive prediction algorithm and a linear regression model. This analysis helps understand the evolution of mental health over time, providing a predictive perspective for future trends. The long-term trend analysis results generated help to find potential health problems and provide references for the establishment of preventive measures.

In the health index monitoring submodule, based on the long-term trend analysis result, a physiological signal monitoring technology is used for real-time monitoring. The method comprises heart rate monitoring, sleep quality analysis and other methods, and abnormal behavior detection and daily activity analysis are combined. Operations in this module include collecting and processing heart rate data, sleep quality metrics, and monitoring daily activities to track health in real time. By continuous monitoring and analysis of these data, signs of health problems can be found in time and responded to in time. The generated health index monitoring result not only provides immediate health information for individuals, but also provides important data support for medical professionals.

In the risk prediction sub-module, based on the health index monitoring result, a machine learning technology and a risk assessment model are applied to predict and assess health risks. This includes nonlinear regression analysis and neural network prediction methods, as well as probabilistic risk scoring and impact factor weight analysis. By these complex analysis methods, the health risk of an individual can be accurately predicted and the importance of various influencing factors can be assessed. The risk prediction not only can warn about upcoming health problems, but also provides scientific basis for taking preventive measures. The risk prediction results generated provide important information to individuals and healthcare providers, which can help make more informed health decisions.

The present invention is not limited to the above embodiments, and any equivalent embodiments which can be changed or modified by the technical disclosure described above can be applied to other fields, but any simple modification, equivalent changes and modification made to the above embodiments according to the technical matter of the present invention will still fall within the scope of the technical disclosure.

Claims (10)

1. Teenager mental health data analysis early warning system, its characterized in that: the system comprises a mood recognition module, a psychological portrait construction module, a risk layering module, a mood prediction module, a behavior analysis module, an intervention strategy module, an early warning feedback module and a health trend monitoring module;

the emotion recognition module is used for carrying out word boundary recognition and emotion tendency analysis by adopting a natural language processing technology based on text, voice and facial expression data, carrying out facial expression feature extraction and voice emotion feature recognition by adopting a deep learning technology, integrating multi-modal information by an integrated model, and generating an emotion recognition result by parameter tuning and cross verification accuracy after model training;

The psychological portrait construction module is used for carrying out homogeneous group recognition by adopting cluster analysis based on emotion recognition results, carrying out feature dimension reduction by adopting principal component analysis, synthesizing individual historical data and behavior features in portrait construction, and generating a psychological portrait by selecting and matching differential group features through an optimization algorithm;

The risk layering module is used for carrying out individual risk division by adopting a support vector machine based on psychological representation, adopting decision tree group risk classification, and then adopting grid search optimization super-parameters in a classification model to generate a risk grade result;

The emotion prediction module processes time series data through an autoregressive moving average model based on historical data and risk grade results, carries out trend and seasonal decomposition, and captures long-term dependency relations through a circulating neural network to generate emotion trend prediction;

The behavior analysis module performs pattern discovery and association rule learning based on social media and learning behavior data by a data mining technology, and graph theory analysis reveals a relationship network structure and generates a behavior analysis result;

The intervention strategy module provides decision support based on a behavior analysis result and emotion trend prediction and an expert system, the rule engine executes a strategy based on set conditions, and the strategy generation process refers to personalized requirements and real-time data updating to generate an intervention strategy;

the early warning feedback module integrates feedback information based on a risk level result and an intervention strategy by adopting multivariate data processing, performs effect evaluation and user response based on a feedback mechanism, and generates early warning notification and feedback by adopting a data-driven decision-making cycle to ensure updating timeliness and early warning accuracy;

the health trend monitoring module is used for identifying abnormal and predicted risks through pattern recognition based on long-term data and trend analysis to reveal a change pattern, carrying out long-term monitoring and model updating, and generating a health trend result based on comprehensive evaluation of health indexes.

2. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the emotion recognition results comprise emotion state distribution, emotion intensity analysis and emotion type recognition, the psychological portrait comprises individual behavior characteristics, group behavior trends and psychological state dynamics, the risk level results comprise high risk, medium risk and low risk classifications, the emotion trend prediction comprises short-term emotion change, long-term emotion trends and potential emotion risks, the behavior analysis results comprise social interaction modes, learning habit analysis and daily activity preference, the early warning notification and feedback comprise early warning message distribution, user feedback collection and early warning effect analysis, and the health trend results comprise long-term psychological health change, key health indexes and potential risk prediction.

3. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the emotion recognition module comprises a text emotion analysis sub-module, a voice emotion analysis sub-module and a facial expression analysis sub-module;

The text emotion analysis submodule is used for carrying out text emotion analysis by adopting a natural language processing technology based on text data, receiving user text through a text crawler or direct input in a data input stage, carrying out part-of-speech tagging, syntactic analysis and emotion vocabulary library matching in a processing stage, carrying out emotion classification and strength evaluation by using a machine learning algorithm, and generating a text emotion analysis result;

The voice emotion analysis submodule adopts a deep learning technology to carry out emotion analysis on voice data, the voice data is input through a microphone or a recording file, feature extraction comprises voice frequency, rhythm and tone analysis, and a convolutional neural network and a cyclic neural network are used for carrying out emotion state recognition to generate a voice emotion analysis result;

The facial expression analysis submodule analyzes facial expression data by adopting an image processing technology, the facial expression data is input through a camera or an image file, the analysis process comprises facial feature point recognition and expression dynamic analysis, and emotion judgment is carried out by utilizing a statistical model and a machine learning method to generate a facial expression emotion analysis result.

4. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the psychological portrait construction module comprises an individual portrait sub-module, a group portrait sub-module and a behavioral pattern analysis sub-module;

The individual portrait sub-module analyzes the emotion and behavior modes of the individual by adopting a clustering analysis method based on the emotion recognition result, specifically adopts a K-means algorithm and a hierarchical clustering method, improves the clustering effect by iteratively optimizing a clustering center and a hierarchical structure, determines the optimal clustering quantity based on a contour coefficient index, and generates an individual psychological portrait;

The group portrait submodule adopts a principal component analysis method to conduct group feature analysis based on the individual psychological portraits, specifically conducts linear dimension reduction to extract principal variables and key features in group data, selects principal component number maximization information retention, evaluates the dimension selection effect through accumulated contribution rate, and generates the group psychological portraits;

The behavior pattern analysis submodule is used for carrying out deep individual and group behavior analysis through a behavior pattern recognition technology based on the group psychological portrait, the analysis comprises time sequence analysis and association rule mining, the behavior trend is predicted based on the time sequence analysis, the association rule mining discovers potential relations among behaviors, and a behavior pattern analysis result is generated.

5. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the risk layering module comprises a risk factor identification sub-module, a risk grade classification sub-module and a risk early warning sub-module;

The risk factor identification sub-module is based on psychological representation, adopts a support vector machine algorithm, performs kernel function conversion to perform high-dimensional mapping, then performs maximum interval classification to identify and analyze risk factors, and simultaneously performs individual behavior and psychological characteristic analysis by using a data mining technology to generate a risk factor identification result;

The risk level classification submodule adopts a decision tree classification technology based on the risk factor identification result, generates and branches and clips by using classification rules, and combines information gain and a keni unrepeace index to carry out risk level classification to generate a risk level classification result;

the risk early warning sub-module is used for carrying out risk monitoring by adopting a correlation analysis method based on a risk grade classification result, collecting and arranging time sequence data, carrying out correlation analysis by adopting a pearson correlation coefficient and a spearman grade correlation, and continuously monitoring risk changes by combining trend analysis and anomaly detection technology to generate a risk early warning result.

6. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the emotion prediction module comprises a time sequence analysis sub-module, a trend prediction sub-module and a prediction result analysis sub-module;

The time sequence analysis submodule performs data import based on historical data, performs emotion change trend analysis by adopting an autoregressive integral moving average model, comprises difference, parameter estimation and model diagnosis, and combines seasonal adjustment and outlier correction to generate a time sequence analysis result;

The trend prediction submodule predicts the emotion trend in the future based on the time sequence analysis result and the risk level result by adopting a circulating neural network and based on a long-short-period memory network and a gating circulating unit, and comprises network structure definition, parameter initialization and iterative training, predicts by utilizing history and current data, and generates an emotion trend prediction result;

The prediction result analysis submodule is used for carrying out data analysis by using a data analysis and pattern recognition technology including a statistical analysis and data visualization method based on the emotion trend prediction result, adopting cluster analysis and association rule mining to identify and refine key emotion driving factors, and carrying out evaluation and optimization to refine analysis results so as to generate a prediction analysis result.

7. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the behavior analysis module comprises a social behavior analysis sub-module, a learning behavior analysis sub-module and a daily behavior analysis sub-module;

The social behavior analysis submodule adopts K-means clustering to conduct feature vectorization on users based on social media data, iteratively optimizes a cluster center to form a user group, then applies a frequent pattern mining algorithm to conduct item set generation and support calculation, identifies a key social behavior pattern and generates a social behavior analysis result;

the learning behavior analysis sub-module analyzes the learning behavior sequence by adopting sequential pattern mining based on the social behavior analysis result, discovers a repeated pattern or trend, determines key learning behavior characteristics by combining decision tree analysis, performs attribute selection and tree construction, classifies and understands differentiated learning patterns, and generates a learning behavior analysis result;

The daily behavior analysis submodule measures the position and the action of an individual in a social network by adopting network centrality analysis based on daily life data, performs shortest path calculation and path dependence analysis by using a path analysis method, reveals the rule and the mode of the daily behavior of the individual, and generates a daily behavior analysis result.

8. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the intervention strategy module comprises an individuation intervention sub-module, a group intervention strategy sub-module and an intervention effect evaluation sub-module;

The individualized intervention sub-module builds a rule base and sets an inference mechanism based on a behavior analysis result and an emotion trend prediction result by adopting a rule-based expert system, and performs emotion and behavior pattern analysis by combining with fuzzy logic, including definition of a fuzzy set and evaluation of fuzzy rules, so as to generate an individualized intervention strategy;

The group intervention strategy sub-module integrates prediction results of a plurality of models by adopting a random forest and gradient lifting mechanism based on an individuation intervention strategy, simultaneously applies social network analysis to analyze network structures and roles, identifies key social dynamics and group behavior modes and generates a group intervention strategy;

the intervention effect evaluation submodule carries out regression analysis of intervention effect based on a group intervention strategy, comprises linear and nonlinear models, matches multiple types of data characteristics and complexity of the intervention effect, carries out comprehensive evaluation by combining objective scores and subjective feedback, analyzes influence of intervention, and generates intervention effect evaluation.

9. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the early warning feedback module comprises an early warning notification sub-module, a feedback collection sub-module and an early warning effect analysis sub-module;

The early warning notification sub-module uses a data-driven notification mechanism based on a risk level result and an intervention strategy, comprises customizing message content and a user behavior tracking algorithm, adopts a behavior pattern recognition technology, recognizes a key behavior pattern, optimizes an information sending strategy, automatically sends early warning information and tracks user response, and generates an early warning notification result;

The feedback collection sub-module collects user feedback through a response type design questionnaire based on an early warning notification result by using an online investigation tool and a feedback analysis algorithm, analyzes the collected data by adopting a text mining and emotion analysis technology, processes the user feedback in real time, and generates a feedback collection result;

The early warning effect analysis submodule adopts a statistical analysis method and an effect evaluation model based on feedback collection results, quantitatively analyzes the collected data through a multivariate analysis and hypothesis testing statistical method, evaluates the effect of intervention measures, comprehensively evaluates the intervention effect by using an effect scoring algorithm and a user satisfaction evaluation model, and generates early warning effect analysis.

10. The system for analyzing and pre-warning teenager mental health data according to claim 1, wherein: the health trend monitoring module comprises a long-term trend analysis sub-module, a health index monitoring sub-module and a risk prediction sub-module;

The long-term trend analysis submodule analyzes historical trend of mental health based on long-term data by adopting a time sequence analysis technology and an adaptive prediction algorithm and a linear regression model, predicts future changes and generates a long-term trend analysis result;

The health index monitoring submodule is used for carrying out real-time monitoring on health indexes by combining abnormal behavior detection and daily activity analysis on the basis of long-term trend analysis results and adopting physiological signal monitoring technologies including heart rate monitoring and sleep quality analysis, and generating health index monitoring results through continuous monitoring and behavior data analysis;

the risk prediction submodule performs health risk prediction and assessment by applying a machine learning technology and a risk assessment model, wherein the machine learning technology and the risk assessment model comprise nonlinear regression analysis and neural network prediction, probability risk scoring and influence factor weight analysis, and generates a risk prediction result.