Mental health data generation method and system based on multi-source data fusion analysis

Abstract

The invention provides a psychological health data generation method based on multi-source data fusion analysis. The method comprises the following steps: acquiring associated data of a user from a plurality of data sources; preprocessing the associated data to obtain preprocessed associated data; performing feature extraction processing and data fusion processing on the preprocessed associated data to obtain a comprehensive data set; training the pre-training model by using the comprehensive data set, and constructing a mental health risk assessment model; and analyzing and processing the user data by using the psychological health risk assessment model to output the psychological health data of the user, the method automatically generates the psychological health data for assessing the psychological health risk of the individual by integrating the associated data of the user from the data source. According to the method, the accuracy and the real-time performance of mental health assessment can be effectively improved, the influence of multi-dimensional factors on mental health is comprehensively considered through data fusion, deviation possibly caused by a single data source is avoided, and the comprehensiveness and the reliability of an assessment result can be ensured.

Description

Technical Field

[0001] The present invention relates to the field of mental health data processing, and particularly to a method and system for generating mental health data based on multi-source data fusion analysis. Background Art

[0002] In recent years, mental health problems have increasingly become the focus of global attention. Especially in the fast-paced modern life, the increasing mental pressure has led to a rising incidence of various mental health problems year by year, and mental health problems have become increasingly prominent. Traditional mental health assessment methods mainly rely on a single data source, such as questionnaires, clinical interviews, etc., lacking comprehensiveness and in-depth analysis, and the accuracy and timeliness of the assessment are limited. These forms rely on the subjective feedback of users, are easily affected by users' emotions and cognitive biases, and the questionnaire data lacks a comprehensive consideration of users' behaviors, physiological states, and environments. In addition, traditional methods take a long time to complete the assessment and cannot achieve real-time monitoring and early warning.

[0003] With the development of Internet of Things, big data, and artificial intelligence technologies, using multi-source data fusion analysis provides new possibilities for mental health risk assessment and has become an important research direction.

[0004] Currently, some studies have attempted to use physiological data, behavioral data, etc. for mental health assessment. However, these methods usually fail to comprehensively utilize the advantages of multiple data sources, fail to fully utilize the complementarity of multi-source data, and lack an effective fusion and analysis mechanism in data processing. There are still many challenges in the integration and analysis of multi-source data in the prior art. Due to low data quality or insufficient feature extraction, the accuracy and generalization ability of the risk assessment model are poor, such as problems like huge data dimensions, serious noise interference, and inconsistent data sources, making it difficult to achieve accurate mental health risk prediction.

[0005] Therefore, there is an urgent need for a new type of mental health risk assessment method to generate mental health data, which can generate more accurate and comprehensive mental health data based on multi-source data fusion analysis as the result of physiological health risk assessment, so as to achieve the purpose of reflecting an individual's mental health status in real time. Summary of the Invention

[0006] The object of the present invention is to address the deficiencies of the prior art and provide a mental health data generation method and system based on multi-source data fusion analysis. The method integrates information from different sources such as physiological data, psychological assessment data, and social media data, automatically evaluates the mental health risks of individuals, can effectively improve the accuracy and real-time performance of mental health assessments, and has broad application prospects. Through data fusion, the method provided by the present invention can comprehensively consider the impact of multi-dimensional factors on mental health, avoid the biases that may be brought by a single data source, and ensure the comprehensiveness and reliability of the evaluation results. And this method is applicable to mental health monitoring and intervention in various fields such as medical, education, and enterprises.

[0007] To achieve the above object, a first aspect of the present invention provides a mental health data generation method based on multi-source data fusion analysis, including:

[0008] Obtaining associated data of a user from multiple data sources;

[0009] Preprocessing the associated data to obtain preprocessed associated data;

[0010] Performing feature extraction processing and data fusion processing on the preprocessed associated data to obtain a comprehensive data set;

[0011] Training a pre-trained model using the comprehensive data set to construct a mental health risk assessment model;

[0012] Analyzing and processing user data using the mental health risk assessment model to output user mental health data.

[0013] Preferably, the associated data includes physiological data, psychological data, social media data, environmental data, and / or behavioral data; where

[0014] The physiological data includes heart rate, blood pressure, skin point response, muscle tension, respiratory rate, and / or hormone level;

[0015] The psychological data includes cognitive function data, emotional state data, personality trait data, and / or stress and coping style data;

[0016] The social media data includes social interaction data, content posting data, and / or privacy settings and account security data;

[0017] The environmental data includes light intensity, noise level, and / or air quality.

[0018] More preferably, the preprocessing of the associated data to obtain preprocessed associated data specifically includes:

[0019] Perform cleaning processing on the associated data to obtain the cleaned associated data; wherein, the cleaning processing includes denoising processing, outlier processing, and / or missing value processing;

[0020] Perform normalization processing on the cleaned associated data to obtain the preprocessed associated data; wherein, the preprocessed associated data includes preprocessed physiological data, preprocessed psychological data, preprocessed social media data, preprocessed environmental data, and preprocessed behavior data.

[0021] Further preferably, the performing feature extraction processing and data fusion processing on the preprocessed associated data to obtain the comprehensive data set specifically includes:

[0022] Perform feature vector extraction processing on the preprocessed physiological data, the preprocessed psychological data, the preprocessed social media data, preprocessed environmental data, and the preprocessed behavior data respectively to obtain a plurality of first feature vectors;

[0023] Concatenate the plurality of first feature vectors to form a high-dimensional feature matrix;

[0024] Adopt a preset decision fusion algorithm to perform weighted fusion on the plurality of first feature vectors of the high-dimensional feature matrix to obtain the first fusion data;

[0025] Use the first feature selection algorithm to perform dimensionality reduction processing on the high-dimensional feature matrix to obtain the comprehensive data set.

[0026] Preferably, the using the comprehensive data set to train a pre-trained model to construct a mental health risk assessment model specifically includes:

[0027] Divide the comprehensive data set into a training set and a test set;

[0028] Use the training set to train the pre-trained model to obtain a pre-trained mental health risk assessment model;

[0029] Use the test set to perform evaluation processing on the pre-trained mental health risk assessment model to obtain model evaluation data;

[0030] Use cross-validation to perform accuracy optimization processing on the pre-trained mental health risk assessment;

[0031] Use hyperparameter tuning to perform generalization ability optimization processing on the pre-trained mental health risk assessment to obtain the mental health risk assessment model.

[0032] Preferably, the user mental health data specifically includes mental health risk level data.

[0033] Further preferably, after outputting the user's mental health data, the method further includes:

[0034] Determine whether the mental health risk level data is a high risk:

[0035] When the mental health risk level data is a high risk, generate a mental health warning message;

[0036] Output the mental health warning message in a first preset output manner.

[0037] Further preferably, the method further includes:

[0038] Generate mental health intervention suggestion data;

[0039] Output the mental health intervention suggestion data in a second preset output manner.

[0040] The second aspect of the present invention provides a mental health data generation system based on multi-source data fusion analysis, including a data acquisition module, a data preprocessing module, a data fusion module, a risk assessment model construction module, and a risk assessment module. Among them,

[0041] The data acquisition module is used to obtain the associated data of the user from multiple data sources;

[0042] The data preprocessing module is used to preprocess the associated data to obtain preprocessed associated data;

[0043] The data fusion module is used to perform feature extraction processing and data fusion processing on the preprocessed associated data to obtain a comprehensive data set;

[0044] The risk assessment model construction module is used to train a pre-trained model using the comprehensive data set to construct a mental health risk assessment model;

[0045] The risk assessment module is used to analyze and process user data using the mental health risk assessment model to output the user's mental health data.

[0046] Further, the system further includes a data output module, which is used to output the mental health data according to a preset output manner.

[0047] In summary, the mental health data generation method and system based on multi-source data fusion analysis provided by the embodiments of the present invention have at least the following beneficial effects:

[0048] 1. Adopt multi-source data fusion to enhance the comprehensiveness of evaluation. By integrating multi-source data, including behavioral data, physiological data, psychological data, social media data, and / or environmental data, as well as various combinations of these data, etc., it can comprehensively reflect the mental health status of users. Behavioral data can reveal the daily behavior patterns of users; physiological data can reflect the physiological stress status of users; psychological data can directly evaluate the psychological state of users; environmental data can reveal the impact of the external environment on users' mental health. Through the fusion analysis of multi-source data, the present invention can more comprehensively capture the mental health risk factors of users, significantly improve the accuracy of evaluating users' mental health, and greatly improve the accuracy and reliability of the generated mental health data.

[0049] 2. Improve the evaluation accuracy. The present invention optimizes the constructed mental health risk assessment model through cross-validation and hyperparameter tuning techniques, further optimizing the model performance and ensuring the generalization ability of the model in different scenarios. Experimental results show that the risk assessment model of the present invention is significantly superior to traditional methods in terms of indicators such as accuracy, recall rate, and F1 score.

[0050] 3. The real-time warning function enhances the timeliness of intervention. The present invention provides a real-time warning function, which can dynamically monitor the mental health status of users and push mental health warning information in a timely manner when high risks are detected. Traditional mental health assessment methods usually take a long time to complete the assessment and are difficult to detect users' mental health problems in a timely manner. By collecting and analyzing multi-source data in real time, the present invention can quickly identify the mental health risks of users and trigger the warning mechanism when the risk level reaches the preset threshold. This real-time warning function can not only help users discover mental health problems in a timely manner, but also provide decision-making support for relevant responsible persons, so as to take intervention measures as early as possible and avoid the problem from deteriorating further.

[0051] 4. The personalized intervention plan enhances the user experience. Traditional mental health intervention methods usually adopt a "one-size-fits-all" approach and are difficult to meet the personalized needs of different users. By analyzing the multi-source data of users, the present invention can accurately identify the mental health problems of users and their causes, and customize an intervention plan for them. This personalized intervention plan can not only improve the intervention effect, but also enhance the sense of participation and satisfaction of users, thus enhancing the overall experience of users.

[0052] 5. Wide range of application scenarios. The present invention has a wide range of application scenarios and is used in various scenarios such as individuals, enterprises, and medical institutions. For individual users, the present invention can help them monitor their mental health status in real time and detect potential problems in a timely manner; for enterprises, the present invention can be used for employee mental health management to improve employees' work efficiency and well-being; for medical institutions, the present invention can assist in mental health diagnosis to improve the diagnosis efficiency and accuracy. In addition, the present invention can be seamlessly integrated with other health management systems to further expand its application scope. The wide range of usage scenarios makes the present invention have high market potential and social value.

[0053] 6. Data-driven scientific decision-making. The present invention provides scientific decision-making support for users and relevant responsible persons in a data-driven manner. Based on multi-source data fusion, the present invention can more objectively and quantitatively evaluate the mental health risks of users and generate visual evaluation reports. It solves the problem that traditional mental health assessment methods usually rely on subjective judgments and are easily affected by human factors. This data-driven decision-making method can not only improve the scientificity and accuracy of decision-making, but also provide strong support for mental health research and policy formulation. For example, medical institutions can use the evaluation results of the present invention to optimize the mental health service process, and government departments can use the data analysis results of the present invention to formulate mental health-related policies. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a flowchart of a method for generating mental health data based on multi-source data fusion analysis provided by an embodiment of the present invention;

[0055] Figure 2 It is a system block diagram of a system for generating mental health data based on multi-source data fusion analysis provided by an embodiment of the present invention;

[0056] Figure 3 It is a flowchart of a method for generating mental health data based on multi-source data fusion analysis of a specific example provided by an embodiment of the present invention;

[0057] Figure 4 It is a schematic diagram of data collection and preprocessing provided by an embodiment of the present invention;

[0058] Figure 5 It is a schematic diagram of constructing a multi-source data fusion and mental health risk assessment model provided by an embodiment of the present invention;

[0059] Figure 6 It is a schematic diagram of a mental health risk assessment module and an application module provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] To make the objectives, technical solutions, and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0061] The technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0062] A method for generating mental health data based on multi-source data fusion analysis provided by an embodiment of the present invention Figure 1 is a flowchart of a method for generating mental health data based on multi-source data fusion analysis provided by an embodiment of the present invention. As Figure 1 shown, the method includes the following steps:

[0063] Step 110: Obtain the associated data of the user from multiple data sources.

[0064] Specifically, the objective of the embodiment of the present invention is to provide a method for generating mental health data based on multi-source data. Therefore, in this step, the data associated with mental health of the user is collected from multiple data sources, that is, the associated data of the user.

[0065] In a preferred solution of the embodiment of the present invention, the associated data includes physiological data, psychological data, social media data, environmental data, and / or behavioral data. Among them,

[0066] the physiological data includes heart rate, blood pressure, skin conductance response, muscle tension, respiratory rate, and / or hormone level. Among them, the change in heart rate can reflect the emotional states of an individual, such as tension and anxiety. For example, when an individual feels tense or anxious, the heart rate may increase. The blood pressure level may be affected by psychological stress, and high stress may lead to an increase in blood pressure. The skin conductance response can indirectly understand the emotional changes of an individual by measuring the electrical conductivity of the skin. For example, the skin conductance may increase when an individual is tense or frightened. The change in muscle tension can reflect the stress state of an individual. For example, the muscles may become tense when an individual is anxious or tense. The change in respiratory rate can reflect the tension and anxiety states of an individual. When an individual feels tense, the breathing may accelerate. The hormone level, such as the cortisol level, usually increases under stress conditions and can be used as a physiological indicator of the stress response.

[0067] The acquisition method of physiological data can be the physiological data of the user obtained through biosensors, such as physiological data such as heart rate, blood pressure, and sleep quality obtained through smart wearable devices such as smart watches, smart bracelets, and smart clothing, as well as medical monitoring devices such as ambulatory blood pressure monitors, continuous glucose monitoring systems, and polysomnography monitors.

[0068] Psychological data Psychological data includes cognitive function data, emotional state data, personality trait data, and / or stress and coping style data.

[0069] In a preferred embodiment of the present invention, the cognitive ability data may further include intelligence level data and / or thinking ability data. The emotional state data may further include emotional stability data and / or emotional experience data. The personality trait data may further include personality inventory assessment data and / or behavior pattern and personality association data. The stress and coping style data may further include stress level assessment data and coping style assessment data.

[0070] In a further preferred embodiment of the present invention, the intelligence level data includes IQ test result data, memory test data, and / or attention test data. Among them, the intelligence level data is the score obtained through tests such as the Wechsler Adult Intelligence Scale, which can reflect an individual's abilities in aspects such as verbal comprehension, perceptual reasoning, working memory, and processing speed, and is an important indicator for evaluating cognitive function. For example, through the Wechsler Adult Intelligence Scale, the strengths and weaknesses of an individual in different cognitive fields can be understood. A low IQ score may indicate a defect in cognitive function and further evaluation and intervention are needed. The memory test data also includes short-term memory data obtained through tests such as the digit span test, long-term memory data obtained through story memory tests, and / or deep memory data, etc. The short-term memory data, long-term memory data, and / or deep memory data can help detect memory disorders. The thinking ability data includes logical thinking test data and / or problem-solving ability test data. Among them, the thinking ability data is obtained through tests such as the Raven's Standard Progressive Matrices, which can be used to evaluate an individual's logical thinking, analogical reasoning, abstract thinking, etc. abilities; the problem-solving ability test data is the data obtained through tests such as the Wisconsin Card Sorting Test, which can be used to evaluate the strategies, flexibility, and abstract thinking abilities for reacting to solve problems.

[0071] In a further preferred embodiment of the present invention, the emotional stability data includes emotional scale assessment data and emotional diary record data. Among them, the emotional scale assessment data is obtained through assessment by scales such as the Self-Rating Anxiety Scale and / or the Self-Rating Depression Scale, and is used to evaluate the degree of anxiety and / or the degree of depression; the emotional diary record data is obtained by recording the daily emotional changes, including the type of emotion, such as happiness, sadness, anger, and anxiety, etc., intensity, which ranges from mild to intense, duration, and triggering factors, etc. From the emotional diary record data, the fluctuation pattern of an individual's emotion can be understood, providing clues for psychological intervention. The emotional experience includes emotional questionnaire data and the associated data between physiological reactions and emotions. Among them, the emotional questionnaire data is obtained through scales specific to certain emotions, and can also be obtained through some comprehensive emotional questionnaires, such as the Positive and Negative Affect Schedule, etc. It can measure the experience degree of both positive and negative emotions simultaneously. The emotional questionnaire data can provide data reference for predicting an individual's psychological state data; the associated data between physiological reactions and emotions indirectly reflects the emotional experience through physiological indicators such as skin conductivity and heart rate variability, which can be used as an objective indicator of emotional experience and can be used as reference data for psychological state assessment.

[0072] In a further preferred embodiment of the present invention, the personality scale assessment data includes Big Five Personality Model scale data and / or Minnesota Multiphasic Personality Inventory data. Among them, the Big Five Personality Model scale data is obtained through testing with the Big Five Personality Model scale, including data on five dimensions: openness, conscientiousness, extraversion, agreeableness, and neuroticism. It can help us understand an individual's personality foundation and predict their psychological reactions when facing stress and setbacks, and thus serve as reference data for predicting mental health risks; the Minnesota Multiphasic Personality Inventory is obtained through the Minnesota Multiphasic Personality Inventory assessment tool, which can be used to detect potential personality disorders and tendencies of mental diseases, and thus serve as reference data for predicting mental health risk assessment. The associated data between behavior patterns and personality includes daily behavior observation data and social behavior analysis data. Among them, the daily behavior observation data is the record data of observing an individual's behavior patterns in daily life, which can reflect whether an individual has aggressiveness, withdrawal, impulsiveness, etc. These data can be used as reference data for mental health risk assessment; the social behavior analysis data is obtained by analyzing an individual's behavior performance in social situations, including, for example, social frequency, social roles, and / or the quality of interpersonal relationships, etc., which can be used as reference data for mental health risks.

[0073] In a further preferred solution of the embodiment of the present invention, the stress level assessment data includes stress perception questionnaire data and / or life event stress scale data. Among them, the perceived stress scale data is data obtained by evaluating the stress level of the subject by asking questions such as the degree of stress, controllability, and overload felt by the subject in the past period of time, and it can be used as reference data for mental health risk assessment; the life event stress scale data is data obtained by evaluating the stress level generated by major events in life, such as the death of a loved one, unemployment, divorce, etc. on an individual, and it can be used as reference data for mental health risk assessment. The coping style assessment data includes coping style questionnaire data and / or defense mechanism assessment data. Among them, the coping style questionnaire data is obtained through a questionnaire survey. This data reflects the main coping styles of an individual when facing stress and difficulties, and the coping style questionnaire data can be used as reference data for mental health risk assessment; the defense mechanism assessment data is data obtained by evaluating the defense mechanisms unconsciously adopted by an individual when facing setbacks and conflicts, which can reflect mental health problems and thus can be used as reference data for mental health risk assessment.

[0074] Social media data may include data such as the emotional state and behavior patterns of users obtained through social media analysis.

[0075] In the preferred solution of the embodiment of the present invention, the social media data includes social interaction data, content publishing data, and / or privacy settings and account security data. Further, the social interaction data includes the number of friends and social density data and / or group participation data; the content publishing data includes the publishing frequency and time and / or the content subject and emotional tendency; the privacy settings and account security data includes the privacy setting level and / or account security behavior.

[0076] In a further preferred embodiment of the present invention, the number of friends and social density include the change data of the number of friends and / or the analysis data of social density. Among them, the change data of the number of friends is the increasing or decreasing trend of the number of friends on the social platform, which can reflect an individual's social activity and mental state. For example, a significant increase in the number of friends within a certain period may indicate that the individual is in an active social state or expanding their social circle. On the contrary, a continuous decrease in the number of friends may imply that the individual is experiencing social withdrawal, which may be related to psychological problems such as depression and social anxiety. Therefore, the change data of the number of friends can be used as reference data for mental health risk assessment; the analysis data of social density is to calculate the interaction frequency with different friends within a certain period (such as the number of chats, the number of likes and comments, etc.) to evaluate social density. A higher social density usually means that the individual has close connections with friends and can obtain more social support, which is beneficial to mental health; while a lower social density may indicate that the individual is relatively isolated in social interactions and is prone to negative emotions such as loneliness. Therefore, the analysis data of social density can be used as reference data for mental health risk assessment. The data on group participation includes the types and quantities of groups joined and / or the speaking and interaction frequencies in the groups. Among them, the types and quantities of groups joined reflect an individual's interests and needs. For example, a person suffering from depression may join a mutual aid group for depression patients to obtain support and understanding from the group. At the same time, the quantity of groups joined can also reflect the degree of the individual's desire for social interaction. Therefore, it can be used as reference data for mental health risk assessment; the speaking and interaction frequencies in the groups refer to actively speaking and participating in discussions in the groups, indicating that the individual is willing to share their views and experiences with others, which may be a manifestation of better mental health. Rarely speaking or staying silent for a long time may mean being relatively passive in social interactions or having some psychological burdens that lead to unwillingness to communicate. Therefore, the speaking and interaction frequencies in the groups can be used as reference data for mental health risk assessment.

[0077] In a further preferred embodiment of the present invention, the release frequency and time include the frequency of released content and / or the release time distribution. Among them, the frequency of released content may reflect that an individual has a strong desire to express and a need to confide, may be a hope to gain the attention and support of others, etc., or may also reflect problems such as unstable individual emotions or anxiety. It can be used as reference data for mental health risk assessment; the release time distribution can reflect an individual's living pattern and emotional state, and may also imply that an individual has insomnia problems or more volatile emotions at night, which may be related to psychological conditions such as psychological stress, depression, and anxiety. Therefore, the release time distribution can also be used as reference data for mental health risk assessment. The content theme and emotional tendency include theme classification and sentiment analysis data. Among them, the theme classification can reflect an individual's trivial matters in life, work and study, emotional relationships, hobbies, etc., and may indicate the psychological state. Therefore, the theme classification can be used as reference data for mental health risk assessment; the sentiment analysis data is to use sentiment analysis tools to judge the emotional tendency of the released content, whether it is positive, negative or neutral. A large amount of negative emotional content (such as sadness, anger, anxiety, etc.) may indicate that an individual is in emotional distress, while more positive emotional content usually reflects a better psychological state. Therefore, the sentiment analysis data can be used as reference data for mental health risk assessment.

[0078] In a further preferred embodiment of the present invention, the privacy setting level includes high privacy setting and / or low privacy setting, which can reflect an individual's sensitivity and also the awareness of privacy protection. Therefore, the privacy setting level can be used as reference data for mental health risk assessment. The account security behavior includes the frequency of password modification and / or the abnormal login situation of the account. The account security behavior may reflect an individual's degree of concern about account security. Therefore, the account security behavior can be used as reference data for mental health risk assessment.

[0079] The environmental data may include data related to the external environment where the user is located, such as the light intensity, noise level, air quality, etc. collected by environmental sensors. The behavior data may include exercise behavior data such as the number of exercise steps, exercise distance, exercise mode, and exercise posture collected by smart devices, as well as daily activity behavior data such as screen usage time, key operation records, application usage frequency and duration, etc. In the embodiments of the present invention, the associated data is not only a simple combination of physiological data, psychological data, social media data, environmental data, and behavior data, but can also be any combination of one or several of these data. The combination method of the data is determined by the data sources that can be obtained.

[0080] In the embodiments of the present invention, the specific types of the biosensor, the environmental sensor, and the intelligent device and their data acquisition methods are not limited. In the prior art, various biosensors, environmental sensors, and intelligent devices that can be obtained are included in the protection scope of the present invention. The associated data obtained by the present invention can be adjusted according to different data types collected by various biosensors, environmental sensors, and intelligent devices in the prior art.

[0081] In the embodiments of the present invention, the collected associated data is transmitted to the data acquisition module of the present invention in real time through the API interface and stored in the data storage platform of the present invention. The data acquisition module of the present invention obtains the associated data while ensuring real-time performance and accuracy. This real-time acquisition method provides high-quality data support for subsequent data preprocessing and fusion analysis.

[0082] Step 120: Preprocess the associated data to obtain preprocessed associated data.

[0083] Specifically, in the embodiments of the present invention, preprocessing the associated data includes data cleaning processing and normalization processing to ensure the quality and consistency of the data. In a specific example of the embodiments of the present invention, this step specifically includes the following steps:

[0084] Step 1201: Clean the associated data to obtain the cleaned associated data. Among them, the cleaning processing includes denoising processing, outlier processing, and / or missing value processing. Any cleaning method that can be used in the prior art can be adopted in this step, such as denoising processing, outlier processing, and / or missing value processing methods to clean the associated data. The specific method selected here is selected according to needs. It is not to perform denoising processing, outlier processing, and missing value processing on each piece of data, but to perform denoising processing on the data that needs denoising processing, outlier processing on the data that needs outlier processing, etc. And for a piece of data, multiple cleaning processing methods can be selected for processing at the same time. For example, physiological data is sequentially subjected to denoising processing, outlier processing, and missing value processing. In a specific example of the embodiments of the present invention, for the abnormal fluctuation of heart rate data, it can be smoothed through a filtering algorithm. In another specific example of the embodiments of the present invention, for the missing data, it can be completed by interpolation. The associated data obtained through the cleaning processing in this step has the characteristics of integrity, consistency, and accuracy.

[0085] Step 1202: Normalize the cleaned associated data to obtain preprocessed associated data. The preprocessed associated data includes preprocessed physiological data, preprocessed psychological data, preprocessed social media data, and / or preprocessed behavioral data. In this step, use the data normalization method in the prior art to normalize the cleaned associated data to convert different data into a unified scale to ensure effective and consistent data for subsequent analysis.

[0086] Step 130: Perform feature extraction processing and data fusion processing on the preprocessed associated data to obtain a comprehensive data set.

[0087] Specifically, in this step, by performing feature extraction processing and data fusion processing on the preprocessed associated data, the multi-source data is integrated to generate a comprehensive data set, which provides a training data basis for training the pre-trained model to construct the mental health risk assessment model provided by the embodiments of the present invention. In the preferred solution of the embodiments of the present invention, this step specifically includes the following steps:

[0088] Step 1301: Respectively perform feature vector extraction processing on the preprocessed physiological data, preprocessed psychological data, preprocessed social media data, and / or preprocessed behavioral data to obtain a plurality of first feature vectors. Specifically, for the associated data of the user obtained in the embodiments of the present invention, select meaningful features from the corresponding preprocessed physiological data, preprocessed psychological data, preprocessed social media data, and / or preprocessed behavioral data after data preprocessing, and perform feature vector extraction according to the selected features to obtain a plurality of first feature vectors. For example, sleep feature vectors and heart rate feature vectors are extracted from the preprocessed physiological data, and emotion feature vectors are extracted from the preprocessed psychological data. More detailed extraction includes, for example, extracting heart rate-related feature vectors, electrocardiogram waveform morphology feature vectors, and blood pressure feature vectors from the preprocessed physiological data; extracting scale score feature vectors, factor analysis feature vectors, simple reaction time feature vectors, choice reaction time feature vectors, eye movement tracking data feature vectors, etc. from the preprocessed psychological data; extracting intensity value feature vectors and light change feature vectors from the light intensity of the preprocessed environmental data, extracting sound pressure level feature vectors and noise spectrum feature vectors from the noise level of the preprocessed environmental data, and extracting pollutant concentration feature vectors and air quality index feature vectors from the air quality of the preprocessed environmental data. Each of these extracted feature vectors is regarded as a first feature vector, that is, a plurality of first feature vectors are extracted in this step.

[0089] Step 1302: Concatenate multiple first feature vectors to form a high-dimensional feature matrix. Specifically, adopt the feature-level fusion technology to concatenate multiple first feature vectors to form a high-dimensional feature matrix. For example, concatenate the sleep feature vector, heart rate feature vector, emotion feature vector, and environmental feature vector into a high-dimensional feature matrix.

[0090] Step 1303: Use a preset decision fusion algorithm to perform weighted fusion on multiple first feature vectors of the high-dimensional feature matrix to obtain first fusion data. Specifically, utilize the decision and fusion technology to perform weighted fusion on multiple first feature vectors of the high-dimensional feature matrix to obtain first fusion data.

[0091] Step 1304: Use the first feature selection algorithm to perform dimensionality reduction on the high-dimensional feature matrix to obtain a comprehensive data set. Specifically, use the principal component analysis feature selection algorithm to perform dimensionality reduction on the fused first fusion data, extract the most important feature vectors for mental health risk assessment, and obtain a comprehensive data set.

[0092] The comprehensive data set obtained after the above steps 1301 to 1304 can be used to train, evaluate, and optimize the pre-trained model selected in the present invention to construct a mental health risk assessment model.

[0093] Step 140: Use the comprehensive data set to train the pre-trained model to construct a mental health risk assessment model.

[0094] Specifically, in this step, based on the comprehensive data set, select a pre-trained model to construct a mental health risk assessment model, and through cross-validation and model optimization, to construct an efficient and accurate risk assessment to construct a mental health assessment model. In the preferred solution of the embodiment of the present invention, constructing a mental health risk assessment model specifically includes the following steps:

[0095] Step 1401: Divide the comprehensive data set into a training set and a test set. Specifically, randomly select a part of the data from the comprehensive data set as the training set, and the remaining data as the test set.

[0096] Step 1402: Use the training set to train the pre-trained model to obtain a pre-trained mental health risk assessment model.

[0097] Specifically, in the embodiments of the present invention, a pre-trained model is selected from the training models in the prior art. In a preferred solution of the embodiments of the present invention, a causal language model that focuses on capturing causal relationships in text is selected as the pre-trained model. For example: In a specific example, qwen series pre-trained models, qwen2.5-7B pre-trained models, etc. are selected as the pre-trained models, and they are trained by the divided training set to obtain a pre-trained mental health risk assessment model. The pre-trained model adjusts its internal parameters by learning the data patterns and feature relationships in the training set to achieve prediction or classification tasks for unknown data.

[0098] Step 1403, use the test set to evaluate the pre-trained mental health risk assessment model to obtain model evaluation data. Specifically, the test set is used to evaluate the performance of the trained model. Specifically, apply the pre-trained mental health risk assessment on the test set, calculate the evaluation metrics for the pre-trained mental health risk assessment model, and obtain the model evaluation data. In a specific example of the embodiments of the present invention, the model evaluation data includes accuracy, recall rate, mean square error, etc.

[0099] Step 1404, use cross-validation to optimize the accuracy of the pre-trained mental health risk assessment. Specifically, based on the model evaluation data obtained in step 1403, use the cross-validation method to optimize the accuracy of the pre-trained mental health risk assessment model to improve the accuracy of the pre-trained mental health risk assessment model.

[0100] Step 1405, use hyperparameter tuning to optimize the generalization ability of the pre-trained mental health risk assessment to obtain a mental health risk assessment model.

[0101] The mental health risk assessment model constructed through steps 1401 to 1405 is applied to the mental health data generation method of the present invention.

[0102] Step 150, use the mental health risk assessment model to analyze and process user data to output user mental health data.

[0103] Specifically, the user data is the historical data of the user, which specifically includes obtaining the historical data of the user from multiple data sources, such as smart wearable devices, medical monitoring devices, etc., and data such as blood pressure and sleep quality can be obtained. It can also be real-time collected user data. Taking the user data as the input of the mental health risk assessment model, after being analyzed and processed by the mental health risk assessment model, user mental health data is output.

[0104] In a preferred solution of the embodiments of the present invention, the user mental health data includes specific mental health risk level data, for example: low risk, medium risk, and high risk.

[0105] In a further preferred solution of the embodiment of the present invention, the mental health risk level data can also be judged. When the mental health risk level data is at a high risk, a mental health warning message is generated, such as "Your mental health level is at a high risk. Please pay attention to keeping a good mood." In the embodiment of the present invention, the generated mental health warning message is also output in a first preset output manner. For example, it is output by voice broadcast or by display on a display screen connected to the system.

[0106] In a further preferred solution of the embodiment of the present invention, the method further includes: generating mental health intervention recommendation data. One way: The system automatically pushes personalized mental health intervention recommendations, such as psychological counseling, relaxation training, etc. Another way, for users with poor sleep quality, the system can recommend relaxation training or sleep improvement suggestions; for users with low mood, the system can recommend psychological counseling or mindfulness practice. And the present invention outputs the mental health intervention recommendations in a second preset output manner, for example, by voice broadcast or by displaying in a preset template on a display screen.

[0107] To more clearly describe a mental health data generation method based on multi-source data fusion analysis provided by the present invention, in a specific example of the embodiment of the present invention, the associated data includes physiological data, psychological data, social media data, and behavioral data. In this specific example, the process of generating mental health data is as Figure 3 shown. In this specific example, the mental health data generation method includes steps 310 to step 350, where steps 320 to 350 correspond to the above-introduced steps 120 to 150. Step 310 is replaced by steps 3101 - 3104, and the specific content executed in each step includes:

[0108] In step 3101, physiological data such as heart rate, blood pressure, and sleep quality are obtained through smart wearable devices such as smart watches, smart bracelets, and smart clothing, as well as medical monitoring devices such as dynamic blood pressure monitors, continuous glucose monitoring systems, and polysomnography monitors.

[0109] In step 3102, the mental health status data of the user is obtained from psychological assessment questionnaires and voice emotion analysis.

[0110] In step 3103, data such as the user's emotional state and behavior pattern are obtained through social media analysis.

[0111] In step 3104, exercise behavior data such as exercise steps, exercise distance, exercise mode, and exercise posture collected by smart devices, as well as daily activity behavior data such as screen usage time, key operation records, application usage frequency, and duration.

[0112] For the convenience of understanding the data acquisition and data preprocessing of the present invention, in another specific example of the embodiments of the present invention, the associated data includes behavioral data, physiological data, psychological data, and environmental data. Figure 4 It is a schematic diagram of data collection and preprocessing provided by the embodiments of the present invention, which shows the data acquisition in step 410 and the data preprocessing in step 420 in a specific embodiment provided by the embodiments of the present invention. Among them, in the data preprocessing stage, it also includes step 4203 for feature extraction of the data. The specific contents executed in step 410 and step 420 are the same as those in step 110 and step 120, and will not be elaborated here.

[0113] For the convenience of understanding the data fusion processing and the mental health risk assessment model in a mental health data generation method based on multi-source data fusion analysis provided by the present invention, a specific example of the embodiments of the present invention is used to illustrate the processing flow. Figure 5 It is a schematic diagram of the construction of multi-source data fusion and mental health risk assessment model provided by the embodiments of the present invention. The specific contents executed in each step include:

[0114] In step 510, each data in the preprocessed associated data is subjected to fusion processing to form a comprehensive data set for training the pre-trained model.

[0115] In step 520, a pre-trained model is selected, and the pre-trained model is trained, optimized, and evaluated using the comprehensive data set, and finally the mental health risk assessment model of the present invention is constructed.

[0116] In step 530, the constructed mental health risk assessment model is used to generate a mental health data generation method based on multi-source data fusion analysis provided by the present invention, as a model for evaluating the mental health risk of users, and is deployed in corresponding application scenarios. For example, it is deployed in various scenarios such as individuals, enterprises, and medical institutions. For individual users, it is used to help them monitor their mental health status in real time and discover potential problems in a timely manner; for enterprises, it is used for employee mental health management to improve work efficiency and happiness of employees; for medical institutions, it is used to assist mental health diagnosis to improve the diagnosis efficiency and accuracy. Figure 6 It is a schematic diagram of the mental health risk assessment module and application module provided by the embodiments of the present invention. As Figure 6 shown, the application process of the mental health risk assessment module constructed by the embodiments of the present invention in a specific scenario includes:

[0117] In step 610, the user data is input into the mental health risk assessment model, and through the prediction of the mental health risk model, mental health data is generated. Among them, it includes dividing the mental health risk level to generate mental health risk level data.

[0118] In step 620, the application module generates a mental health report based on the mental health data generated by the mental health risk assessment module. It can further generate mental health early warning information and mental health intervention recommendation data according to the mental health risk level, and output them in a preset manner to provide mental health reference for users.

[0119] The above introduces a method for generating mental health data based on multi-source data fusion analysis provided by the present invention. Next, a system for implementing the method for generating mental health data based on multi-source data fusion analysis provided by the present invention will be introduced. Figure 2 As shown in the system block diagram of a system for generating mental health data based on multi-source data fusion analysis provided by the present invention, a system 1000 for generating mental health data based on multi-source data fusion analysis provided by an embodiment of the present invention includes: a data acquisition module 1001, a data preprocessing module 1002, a data fusion module 1003, a risk assessment model construction module 1004, and a risk assessment module 1005.

[0120] The data acquisition module 1001 is used to obtain the associated data of the user from multiple data sources.

[0121] The data preprocessing module 1002 is used to preprocess the associated data to obtain preprocessed associated data.

[0122] The data fusion module 1003 is used to perform feature extraction processing and data fusion processing on the preprocessed associated data to obtain a comprehensive data set.

[0123] The risk assessment model construction module 1004 is used to train a pre-trained model using the comprehensive data set to construct a mental health risk assessment model.

[0124] The risk assessment module 1005 is used to analyze and process the user data using the mental health risk assessment model to output the user's mental health data.

[0125] In another solution of the embodiment of the present invention, the system 1000 for generating mental health data based on multi-source data fusion analysis further includes a data output module 1006, which can be a voice output device or a display output device, etc., and is used to output the mental health data according to a preset output method.

[0126] Each component module of the system for generating mental health data based on multi-source data fusion analysis provided by the embodiment of the present invention is integrated into a complete system while ensuring clear data flow and logical relationship between the modules, and the data transmission between the data acquisition module 1001 and the data preprocessing module 1002 is realized through the API structure. This system uses distributed computing technology to accelerate large-scale data processing and uses model compression technology to reduce the model calculation complexity.

[0127] A mental health data generation method and system based on multi-source data fusion analysis provided by an embodiment of the present invention obtain associated data of a user by acquiring multi-source data of the user, preprocess the data, train based on the associated data to obtain a mental health risk assessment model, and use the constructed mental health risk assessment model to evaluate the collected user data and output the mental health data of the user. The beneficial effects of the present invention are at least reflected in the following aspects:

[0128] 1. Adopting multi-source data fusion to enhance the comprehensiveness of evaluation. By integrating multi-source data, including behavioral data, physiological data, psychological data, environmental data, etc., it can comprehensively reflect the mental health status of users. Behavioral data can reveal the daily behavior patterns of users; physiological data can reflect the physiological stress status of users; psychological data can directly evaluate the mental state of users; environmental data can reveal the impact of the external environment on the mental health of users. Through the fusion analysis of multi-source data, the present invention can more comprehensively capture the mental health risk factors of users, significantly improve the accuracy of evaluating the mental health of users, and greatly improve the accuracy and reliability of the generated mental health data.

[0129] 2. Improving the evaluation accuracy. The present invention optimizes the constructed mental health risk assessment model through cross-validation and hyperparameter tuning techniques to further optimize the model performance and ensure the generalization ability of the model in different scenarios. Experimental results show that the risk assessment model of the present invention is significantly superior to traditional methods in terms of indicators such as accuracy, recall rate, and F1 score.

[0130] 3. The real-time warning function improves the timeliness of intervention. The present invention provides a real-time warning function, which can dynamically monitor the mental health status of users and push mental health warning information in a timely manner when high risks are detected. Traditional mental health assessment methods usually take a long time to complete the assessment and it is difficult to detect the mental health problems of users in a timely manner. By collecting and analyzing multi-source data in real time, the present invention can quickly identify the mental health risks of users and trigger the warning mechanism when the risk level reaches the preset threshold. This real-time warning function can not only help users detect mental health problems in a timely manner, but also provide decision-making support for relevant responsible persons, so as to take intervention measures as early as possible and avoid the further deterioration of problems.

[0131] 4. The personalized intervention plan improves the user experience. Traditional mental health intervention methods usually adopt a "one-size-fits-all" approach and it is difficult to meet the personalized needs of different users. By analyzing the multi-source data of users, the present invention can accurately identify the mental health problems of users and their causes, and customize an intervention plan for them. This personalized intervention plan can not only improve the intervention effect, but also enhance the sense of participation and satisfaction of users, thus improving the overall experience of users.

[0132] 5. Wide range of application scenarios. The present invention has a wide range of application scenarios and is applicable to various scenarios such as individuals, enterprises, and medical institutions. For individual users, the present invention can help them monitor their mental health status in real time and detect potential problems in a timely manner; for enterprises, the present invention can be used for employee mental health management to improve employees' work efficiency and well-being; for medical institutions, the present invention can assist in mental health diagnosis to improve the diagnosis efficiency and accuracy. In addition, the present invention can be seamlessly integrated with other health management systems to further expand its application scope. The wide range of usage scenarios makes the present invention have high market potential and social value.

[0133] 6. Data-driven scientific decision-making. The present invention provides scientific decision-making support for users and relevant responsible persons in a data-driven manner. Based on multi-source data fusion, the present invention can more objectively and quantitatively evaluate the mental health risks of users and generate visual evaluation reports. It solves the problem that traditional mental health assessment methods usually rely on subjective judgment and are easily affected by human factors. This data-driven decision-making method can not only improve the scientificity and accuracy of decision-making, but also provide strong support for mental health research and policy formulation. For example, medical institutions can use the evaluation results of the present invention to optimize the mental health service process, and government departments can use the data analysis results of the present invention to formulate mental health-related policies.

[0134] Professional personnel should also be able to further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0135] The steps of the method or algorithm described in combination with the embodiments disclosed herein can be implemented by hardware, software modules executed by a processor, or a combination of the two. The software modules can be placed in a random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the technical field.

[0136] The specific embodiments described above have further elaborated on the object, technical solution and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for generating mental health data based on multi-source data fusion analysis, characterized in that, The method includes: Obtaining the associated data of the user from multiple data sources; Preprocessing the associated data to obtain preprocessed associated data; Performing feature extraction processing and data fusion processing on the preprocessed associated data to obtain a comprehensive data set; Using the comprehensive data set to train a pre-trained model to construct a mental health risk assessment model; Using the mental health risk assessment model to analyze and process user data to output user mental health data.

2. The mental health data generation method according to claim 1, characterized in that The associated data includes physiological data, psychological data, social media data, environmental data, and / or behavioral data; wherein, The physiological data includes heart rate, blood pressure, skin point response, muscle tension, respiratory rate, and / or hormone level; The psychological data includes cognitive function data, emotional state data, personality trait data, and / or stress and coping style data; The social media data includes social interaction data, content publishing data, and / or privacy settings and account security data; The environmental data includes light intensity, noise level, and / or air quality.

3. The mental health data generation method according to claim 2, characterized in that, The preprocessing of the associated data to obtain preprocessed associated data specifically includes: Performing cleaning processing on the associated data to obtain the cleaned associated data; wherein, the cleaning processing includes denoising processing, outlier processing, and / or missing value processing; Performing normalization processing on the cleaned associated data to obtain the preprocessed associated data; wherein, the preprocessed associated data includes preprocessed physiological data, preprocessed psychological data, preprocessed social media data, preprocessed environmental data, and preprocessed behavioral data.

4. The mental health data generation method according to claim 3, wherein The performing of feature extraction processing and data fusion processing on the preprocessed associated data to obtain a comprehensive data set specifically includes: Performing feature vector extraction processing on the preprocessed physiological data, the preprocessed psychological data, the preprocessed social media data, the preprocessed environmental data, and the preprocessed behavioral data respectively to obtain a plurality of first feature vectors; Concatenating the plurality of first feature vectors to form a high-dimensional feature matrix; Using a preset decision fusion algorithm to perform weighted fusion on the plurality of first feature vectors of the high-dimensional feature matrix to obtain first fusion data; Using a first feature selection algorithm to perform dimensionality reduction processing on the high-dimensional feature matrix to obtain the comprehensive data set.

5. The mental health data generation method according to claim 1, wherein The using of the comprehensive data set to train a pre-trained model to construct a mental health risk assessment model specifically includes: Dividing the comprehensive data set into a training set and a test set; Using the training set to train the pre-trained model to obtain a pre-trained mental health risk assessment model; Using the test set to perform evaluation processing on the pre-trained mental health risk assessment model to obtain model evaluation data; Using cross-validation to perform accuracy optimization processing on the pre-trained mental health risk assessment; Using hyperparameter tuning to perform generalization ability optimization processing on the pre-trained mental health risk assessment to obtain the mental health risk assessment model.

6. The mental health data generation method according to claim 1, characterized in that The user mental health data specifically includes mental health risk level data.

7. The method for generating mental health data according to claim 6, wherein After the output of the user mental health data, the method further includes: Determine whether the mental health risk level data is a high risk: When the mental health risk level data is a high risk, generate a mental health warning message; Output the mental health warning message in a first preset output manner.

8. The mental health data generation method according to claim 7, wherein The method further includes: Generate mental health intervention suggestion data; Output the mental health intervention suggestion data in a second preset output manner.

9. A mental health data generation system based on multi-source data fusion analysis, characterized in that, The system includes a data collection module, a data preprocessing module, a data fusion module, a risk assessment model construction module, and a risk assessment module, where The data collection module is used to obtain the associated data of the user from multiple data sources; The data preprocessing module is used to preprocess the associated data to obtain preprocessed associated data; The data fusion module is used to perform feature extraction processing and data fusion processing on the preprocessed associated data to obtain a comprehensive data set; The risk assessment model construction module is used to train a pre-trained model using the comprehensive data set to construct a mental health risk assessment model; The risk assessment module is used to analyze and process user data using the mental health risk assessment model to output user mental health data.

10. The mental health data generation system according to claim 9, characterized in that, The system further includes a data output module, which is used to output the mental health data according to a preset output manner.

Images