Student mental health early warning method and early warning system based on AI enabling

Abstract

The invention relates to the technical field of mental health early warning, in particular to a student mental health early warning method and early warning system based on AI enabling. Collecting metadata of psychological health activities of students in real time by constructing an event monitor and an event trigger; converting the psychological activity state when the student executes the structured psychological health activity into a difference equation, and solving and processing the psychological activity state difference equation by using the metadata to obtain a psychological activity fluctuation state matrix of the student when each piece of metadata is generated; training a rating model by using a clustering fluctuation state matrix of the associated crisis rating thermodynamic diagram of the psychological health of the student, and generating a psychological health crisis rating model; and mapping a standardized psychological health rating model and a psychological health crisis rating model when the psychology of the student is in a health standard state, and analyzing the distribution distance offset so as to carry out monitoring and early warning. The system can continuously monitor and evaluate psychological states of school students and send out early warning signals, so that corresponding intervention measures are taken.

Description

Technical Field

[0001] The present invention relates to the field of mental health early warning technology, and in particular to a student mental health early warning method and early warning system based on AI empowerment. Background Art

[0002] In recent years, students' mental health problems have become increasingly prominent, and how to identify and intervene in students' mental problems in a timely and effective manner has become an urgent problem to be solved. Traditional mental health early warning methods mainly rely on manual evaluation, which has limitations such as strong subjectivity, low efficiency and limited coverage. This results in a large error rate in the early warning between the evaluation results and the early warning results executed based on the evaluation results, making it difficult for educational managers to better and more accurately obtain reference information on students' mental activities, and thus difficult to take corresponding intervention measures. The rapid development of artificial intelligence technology has made it possible to build a more intelligent and precise student mental health early warning system. Therefore, an AI-enabled intelligent student mental health early warning system can be built for use in schools, educational institutions and other scenarios to conduct comprehensive mental health monitoring and evaluation and early warning of students' mental activities, providing a scientific basis for early detection and timely intervention of students' mental problems. Summary of the invention

[0003] The present invention overcomes the shortcomings of the prior art and provides a student mental health early warning method and early warning system based on AI empowerment.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] The first aspect of the present invention provides a student mental health early warning method based on AI empowerment, comprising the following steps:

[0006] S102: Obtain several structured mental health activities of students' mental health and their event variable rules and necessary triggering behaviors, construct event triggers and event listeners according to the event variable rules and necessary triggering behaviors, collect students' mental health activity data in real time through event listeners and event triggers, and obtain metadata;

[0007] S104: Obtaining the psychological behavior norms of the students' mental health, constructing a psychological activity state differential equation when the students perform structured mental health activities based on the psychological behavior norms, solving the psychological activity state differential equation with metadata, and obtaining a fluctuation state matrix of the students' psychological activities when each metadata is generated;

[0008] S106: drawing a correlation crisis rating heat map based on the correlation crisis rating of the students' mental health, clustering each mental activity fluctuation state of the fluctuation state matrix into the correlation crisis rating heat map, obtaining a clustering result, and using the clustering result to train a rating model to generate a mental health crisis rating model;

[0009] S108: A standardized mental health rating model is fitted through the expected psychological rating range preset when the students' psychology is in a healthy standard state, the mental health crisis rating model is mapped to the standardized mental health rating model and the distribution distance deviation is analyzed to provide early warning for students with mental health abnormalities.

[0010] More specifically, the step S102 includes the following steps:

[0011] Obtaining preset monitoring requirements for students' mental health, and extracting several structured mental health activities required for monitoring to assess students' mental health through the preset monitoring requirements, which are defined as susceptible structured events;

[0012] Obtaining event variable rules for each susceptible structured event, and setting the event type of each susceptible structured event and the necessary triggering behavior corresponding to each event type according to the event variable rules;

[0013] Create an event trigger architecture based on the event type of each susceptible structured event, set the event source according to the necessary trigger behavior corresponding to each event type, and follow the event variable rules to generate the event source's capture decision for each event type;

[0014] Using the capture decision of each event type by the event source, an event trigger architecture is written to obtain an event trigger for each susceptible structured event with necessary triggering behavior;

[0015] Obtain the historical operation logic of the event type corresponding to the necessary triggering behavior operation after the event trigger executes the capture decision, obtain the trigger response function of the necessary triggering behavior in the big data based on the historical operation logic, and build a series of event listeners that can be associated with the event type according to the trigger response function;

[0016] Stripping the system node network of each susceptible structured event, deploying a series of event listeners in the system node network corresponding to the susceptible structured event, and monitoring the real-time event source of each necessary triggering behavior operation corresponding to the event type in the system node network through event triggers to obtain the real-time trigger result of each event trigger;

[0017] When a series of event listeners receive the corresponding real-time trigger results, the hash state functions between the real-time event source and different historical operation logics are calculated one by one, and only the historical operation logic corresponding to the maximum hash state function is extracted and marked as the established operation logic;

[0018] The event listener that receives the real-time trigger result is controlled to execute the predetermined operation logic to obtain multiple metadata of the student's structured mental health activities at the current timestamp.

[0019] More specifically, the step S104 includes the following steps:

[0020] Obtain the psychological behavior norms of students' mental health, and use the psychological behavior norms as the constraint benchmark to obtain the psychological activity flow when structured mental health activities are executed according to event variable rules under the constraint benchmark;

[0021] Introducing the difference method to obtain the differential linear invariant function of the psychological activity process, if the differential linear invariant function is within a preset differential linear invariant function interval, then calculating and constructing the activity state of the psychological activity process in the difference method based on the differential linear invariant function to generate a psychological activity state differential equation;

[0022] The expected state space time series of the current timestamp under the preset differential linear invariant function is used to convert the multiple metadata of the student's structured mental health activities at the current timestamp into state variables at the current timestamp time series based on the expected state space time series, which are defined as mental activity state variables;

[0023] Derivation of a differential relationship of metadata expressing psychological activity fluctuations based on psychological activity state variables, substituting the psychological activity state variables into the psychological activity state differential equation for updating, and generating a derived and updated psychological activity state variable;

[0024] Substituting the updated mental activity state variables derived from the derivation back into the mental activity state difference equation to perform reconstruction processing, thereby obtaining a new mental activity state difference equation;

[0025] Solve the differential equation of the psychological activity state of the freshmen to obtain the state space expression of the change of the psychological activity state when each metadata is generated, and determine the fluctuation state matrix of the student's psychological activity when each metadata is generated during the psychological activity process according to the description of the state space expression.

[0026] More specifically, the step S106 includes the following steps:

[0027] Acquire crisis prevention standards and regulations for students' mental health, and acquire multiple related crisis ratings of students' mental health activities with different structured mental health activities based on the crisis prevention standards and regulations;

[0028] Construct a heat map, retrieve the mental health assessment scores of students when their mental health reaches each associated crisis rating based on the mental health management log, fit the mental health assessment scores corresponding to each associated crisis level in the heat map, and generate an associated crisis rating heat map;

[0029] Obtain the color index of each associated crisis rating through the associated crisis rating heat map, obtain the psychological activity state-color coding knowledge map based on the big data network, use the psychological activity state-color coding knowledge map to identify the fluctuation state matrix, and obtain the color coding sequence of each psychological activity fluctuation state;

[0030] Constructing N sub-cluster clusters based on the color index of each associated crisis rating, calculating the membership of each sub-cluster cluster through the color coding sequence of each psychological activity fluctuation state, and obtaining a number of current membership matrices;

[0031] Calculating the center vector of each sub-cluster according to each current membership matrix to obtain multiple new sub-cluster centers, iteratively updating the membership of each mental activity fluctuation state in the fluctuation state matrix according to the multiple new sub-cluster centers, and outputting the distance offset value of each membership matrix;

[0032] A maximum distance offset threshold is preset, and the above-mentioned update iterative steps are repeated to continuously calculate the distance offset value of the membership matrix. When the distance offset value is less than the maximum distance offset threshold, the iterative update is stopped, and finally a clustering result of each psychological activity fluctuation state in the fluctuation state matrix is ​​generated, which is clustered in each associated crisis rating;

[0033] A crisis rating model is constructed based on the associated crisis rating heat map, and data training and verification are performed on the clustering results in the crisis rating model to obtain a mental health crisis rating model.

[0034] More specifically, the step S108 includes the following steps:

[0035] Obtain the preset expected psychological rating range when the student's psychology is in a healthy standard state, and simultaneously obtain one or more benchmark psychological rating indicators corresponding to the expected psychological rating range for each structured mental health activity;

[0036] Obtaining the established qualified rating scores of one or more benchmark psychological rating indicators according to the expected psychological rating range, and comprehensively estimating the qualified rating standard score threshold of each structured mental health activity through the established qualified rating scores;

[0037] Constructing a qualified rating model, fitting points in the qualified rating model based on the qualified rating standard score thresholds of each structured mental health activity, and obtaining a standardized mental health rating model for students' mental health;

[0038] Creating a rating distribution network of structured mental health activities, extracting the rating distribution network in the mental health crisis rating model, defined as a first rating distribution network, and extracting the rating distribution network of the standardized mental health rating model, defined as a second rating distribution network;

[0039] Preset a mapping cost function based on the expected psychological rating range, map the first rating distribution network to the second rating distribution network with the mapping cost function as the mapping purpose, and obtain an actual joint distribution set of rating distributions;

[0040] Preset a warning threshold when the student's mental health reaches a dangerous rating, establish an abnormal joint distribution set based on the warning threshold, and introduce the Sinkhorn distance method to calculate the distribution offset distance between the actual joint distribution set and the abnormal joint distribution set;

[0041] If the distribution offset distance is greater than the preset distribution offset distance, the student is marked as a normal mental health student and no warning signal is generated;

[0042] If the distribution offset distance is less than the preset distribution offset distance, the student is marked as an abnormally mentally healthy student. At this time, a warning signal with a corresponding rating is generated based on the distribution offset distance to issue a mental health warning.

[0043] The second aspect of the present invention provides a student mental health early warning system based on AI empowerment, the student mental health early warning system includes a memory and a processor, the memory stores a student mental health early warning method program based on AI empowerment, and when the student mental health early warning method program is executed by the processor, any one of the steps of the student mental health early warning method is implemented.

[0044] The third aspect of the present invention provides a computer-readable storage medium, which stores a computer detection program. When the computer detection program is executed by at least one processor, it can implement any one of the AI-enabled student mental health early warning methods.

[0045] The present invention solves the technical defects existing in the background technology, and the beneficial technical effects of the present invention are:

[0046] Obtain several structured mental health activities of students' mental health and their event variable rules and necessary triggering behaviors, construct event triggers and event listeners according to the event variable rules and necessary triggering behaviors, collect students' mental health activity data in real time through event listeners and event triggers, and obtain metadata; obtain the psychological behavior norms of students' mental health, construct the psychological activity state differential equations when students perform structured mental health activities based on the psychological behavior norms, solve and process the psychological activity state differential equations with metadata, and obtain the fluctuation state matrix of students' mental activities when each metadata is generated; draw the correlation crisis rating heat map based on the correlation crisis rating of students' mental health, cluster each psychological activity fluctuation state of the fluctuation state matrix into the correlation crisis rating heat map, obtain the clustering results, train the rating model with the clustering results, and generate the mental health crisis rating model; fit the standardized mental health rating model through the expected psychological rating range preset when the students' psychology is in the healthy standard state, map the mental health crisis rating model to the standardized mental health rating model and analyze the distribution distance offset, so as to warn students with abnormal mental health. On the one hand, the present invention can continuously monitor and evaluate the mental state of school students and send out warning signals, so as to take corresponding intervention measures; on the other hand, it can achieve in-depth understanding and effective management of students' mental health. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0048] Figure 1 A first method flow chart of a student mental health early warning method based on AI empowerment is shown;

[0049] Figure 2 A second method flow chart of a student mental health early warning method based on AI empowerment is shown;

[0050] Figure 3 A system framework diagram of a student mental health early warning system based on AI empowerment is shown. DETAILED DESCRIPTION

[0051] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0052] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0053] The first aspect of the present invention provides a student mental health early warning method based on AI empowerment, such as Figure 1 As shown, the following steps are included:

[0054] S102: Obtain several structured mental health activities of students' mental health and their event variable rules and necessary triggering behaviors, construct event triggers and event listeners according to the event variable rules and necessary triggering behaviors, collect students' mental health activity data in real time through event listeners and event triggers, and obtain metadata;

[0055] S104: Obtaining the psychological behavior norms of the students' mental health, constructing a psychological activity state differential equation when the students perform structured mental health activities based on the psychological behavior norms, solving the psychological activity state differential equation with metadata, and obtaining a fluctuation state matrix of the students' psychological activities when each metadata is generated;

[0056] S106: drawing a correlation crisis rating heat map based on the correlation crisis rating of the students' mental health, clustering each mental activity fluctuation state of the fluctuation state matrix into the correlation crisis rating heat map, obtaining a clustering result, and using the clustering result to train a rating model to generate a mental health crisis rating model;

[0057] S108: A standardized mental health rating model is fitted through the expected psychological rating range preset when the students' psychology is in a healthy standard state, the mental health crisis rating model is mapped to the standardized mental health rating model and the distribution distance deviation is analyzed to provide early warning for students with mental health abnormalities.

[0058] More specifically, the step S102 includes the following steps:

[0059] Obtaining preset monitoring requirements for students' mental health, and extracting several structured mental health activities required for monitoring to assess students' mental health through the preset monitoring requirements, which are defined as susceptible structured events;

[0060] Obtaining event variable rules for each susceptible structured event, and setting the event type of each susceptible structured event and the necessary triggering behavior corresponding to each event type according to the event variable rules;

[0061] Create an event trigger architecture based on the event type of each susceptible structured event, set the event source according to the necessary trigger behavior corresponding to each event type, and follow the event variable rules to generate the event source's capture decision for each event type;

[0062] Using the capture decision of each event type by the event source, an event trigger architecture is written to obtain an event trigger for each susceptible structured event with necessary triggering behavior;

[0063] Obtain the historical operation logic of the event type corresponding to the necessary triggering behavior operation after the event trigger executes the capture decision, obtain the trigger response function of the necessary triggering behavior in the big data based on the historical operation logic, and build a series of event listeners that can be associated with the event type according to the trigger response function;

[0064] Stripping the system node network of each susceptible structured event, deploying a series of event listeners in the system node network corresponding to the susceptible structured event, and monitoring the real-time event source of each necessary triggering behavior operation corresponding to the event type in the system node network through event triggers to obtain the real-time trigger result of each event trigger;

[0065] When a series of event listeners receive the corresponding real-time trigger results, the hash state functions between the real-time event source and different historical operation logics are calculated one by one, and only the historical operation logic corresponding to the maximum hash state function is extracted and marked as the established operation logic;

[0066] The event listener that receives the real-time trigger result is controlled to execute the predetermined operation logic to obtain multiple metadata of the student's structured mental health activities at the current timestamp.

[0067] It should be noted that students' psychological and emotional states are easily affected by different structured activities and fluctuate. For example, students may feel happy when participating in certain social activities or may develop depression when arguing with others. Therefore, it is necessary to collect data to evaluate students' mental health by monitoring the changes in their psychological states when performing these structured activities. Since students always follow some event variable rules when performing these structured activities and their psychological states will change only when necessary triggering behaviors occur, in the above example, when students participate in certain social activities and feel happy, they may play with their bodies after talking to others about some pleasant words and conversations. Their conversations are event variable rules, and physical play is the necessary triggering behavior. However, it is difficult for traditional monitoring and acquisition modules to deeply monitor from the above perspectives, so that the psychological state data collected in the end cannot describe the psychological activities of students during structured activities, resulting in large errors in the modeling of subsequent evaluation and early warning models and intervention feedback, making it difficult to achieve more accurate early warning and intervention effects. To this end, this method constructs event triggers based on the event variable rules and necessary triggering behaviors when students perform various structured mental health activities, and deploys them in the system node network susceptible to structured events, so as to accurately monitor and capture the triggering behaviors that cause changes in the psychological state of students during the structured mental health activities in real time, thereby improving the monitoring reliability of mental health data; and based on the operation logic training when the event is triggered after the event trigger executes the capture decision, a series of event listeners are constructed to respond to the types of events triggered by different aspects of the monitored psychological state. When an event trigger monitors a real-time event source corresponding to an event type of a necessary triggering behavior operation, the real-time event source should be analyzed in response to the different historical operation logics for training the event listener, so as to obtain an operation logic adapted to collect the psychological state change data under the real-time event source, and finally the metadata of the student in the structured mental health activity at the current timestamp can be accurately obtained.

[0068] It should be noted that structured mental health activities include academic examination activities, family life activities, campus activities, social activities and life habit activities; event types include test answers, family life daily life, campus life daily life, social experiences and life habit behaviors; necessary trigger behaviors include test scores, family life trivia, campus life trivia and social negative emotions. This method can build a series of event triggers and event listeners based on the event types and psychological change trigger behaviors when students participate in structured activities, so as to accurately and reliably monitor and collect students' mental health activity metadata at the current timestamp, providing highly credible data support for the construction of subsequent mental health rating models, abnormal warnings of students' mental states and the formulation of intervention plans, ensuring the real-time collection of students' mental health data and the accuracy of mental health warnings.

[0069] More specifically, the step S104 is as follows: Figure 2 As shown, the specific steps include:

[0070] S202: Obtain the psychological behavior norms of students' mental health, and use the psychological behavior norms as the constraint benchmark to obtain the psychological activity flow when the structured mental health activities are executed according to the event variable rules under the constraint benchmark;

[0071] S204: introducing a differential method to obtain a differential linear invariant function of the psychological activity process. If the differential linear invariant function is within a preset differential linear invariant function interval, then calculating and constructing the activity state of the psychological activity process in the differential method based on the differential linear invariant function to generate a psychological activity state differential equation;

[0072] S206: Preset the expected state space time series of the current timestamp under the differential linear invariant function, and based on the expected state space time series, convert multiple metadata of the student in the structured mental health activity at the current timestamp into state variables at the current timestamp time series, and define them as mental activity state variables;

[0073] S208: deriving a differential relationship of metadata expressing psychological activity fluctuations based on the psychological activity state variables, substituting the psychological activity state variables into the psychological activity state differential equation for updating, and generating a derived and updated psychological activity state variable;

[0074] S210: Substituting the psychological activity state variables updated by the derivation back into the psychological activity state difference equation to perform reconstruction processing, and obtaining a new psychological activity state difference equation;

[0075] S212: Solve the differential equation of the psychological activity state of the freshmen to obtain the state space expression of the change of the psychological activity state when each metadata is generated, and determine the fluctuation state matrix of the student's psychological activity when each metadata is generated during the psychological activity process according to the description of the state space expression.

[0076] It should be noted that the metadata collected by event triggers and time listeners are discrete data at the current timestamp, and these discrete metadata are difficult to express the fluctuations in students' mental state during their participation in structured mental health activities, and increase the difficulty of constructing subsequent linear rating and early warning models. At the same time, unnecessary calculation steps are added, which may reduce the accuracy of the constructed rating and early warning model. Therefore, this method further streamlines the process of students participating in structured mental health activities to obtain the mental activity process. The purpose of streamlining is to make the psychological state expression of metadata maintain the linear specification when the structured mental health activity is executed according to the event variable rule. However, at this time, it cannot be directly differentiated according to the mental activity process. It is necessary to further judge whether it has certain linear requirements. Therefore, it is judged whether the differential linear invariant function of the mental activity process is in the preset differential linear invariant function interval. If it is, it means that the mental activity process maintains linear differential invariance, otherwise it means that it does not maintain; then, based on the mental activity process and the differential linear invariant function, a mental activity state differential equation is constructed. The equation describes an original state of the mental activity in the mental activity process, and then the equation is derived and solved with metadata as the state variable, so as to update this original state according to the fluctuation of the mental activity expressed by the metadata, so as to finally generate a fluctuation state matrix to reflect the fluctuation of the mental activity of students when each metadata is generated in the process of the mental activity process, so that the metadata is no longer discretized and maintains a certain linearity. This method can use the psychological activity process of students when participating in structured activities as the linear basis of the psychological activity state, and use metadata differential deduction to describe the fluctuations of the psychological activity state, so that the metadata maintains temporal linearity, and the metadata can accurately describe the fluctuations of the students' psychological activity state, improve the accuracy and reliability of subsequent model construction, and reduce unnecessary model construction operation steps.

[0077] More specifically, the step S106 includes the following steps:

[0078] Acquire crisis prevention standards and regulations for students' mental health, and acquire multiple related crisis ratings of students' mental health activities with different structured mental health activities based on the crisis prevention standards and regulations;

[0079] Construct a heat map, retrieve the mental health assessment scores of students when their mental health reaches each associated crisis rating based on the mental health management log, fit the mental health assessment scores corresponding to each associated crisis level in the heat map, and generate an associated crisis rating heat map;

[0080] Obtain the color index of each associated crisis rating through the associated crisis rating heat map, obtain the psychological activity state-color coding knowledge map based on the big data network, use the psychological activity state-color coding knowledge map to identify the fluctuation state matrix, and obtain the color coding sequence of each psychological activity fluctuation state;

[0081] Constructing N sub-cluster clusters based on the color index of each associated crisis rating, calculating the membership of each sub-cluster cluster through the color coding sequence of each psychological activity fluctuation state, and obtaining a number of current membership matrices;

[0082] Calculating the center vector of each sub-cluster according to each current membership matrix to obtain multiple new sub-cluster centers, iteratively updating the membership of each mental activity fluctuation state in the fluctuation state matrix according to the multiple new sub-cluster centers, and outputting the distance offset value of each membership matrix;

[0083] A maximum distance offset threshold is preset, and the above-mentioned update iterative steps are repeated to continuously calculate the distance offset value of the membership matrix. When the distance offset value is less than the maximum distance offset threshold, the iterative update is stopped, and finally a clustering result of each psychological activity fluctuation state in the fluctuation state matrix is ​​generated, which is clustered in each associated crisis rating;

[0084] A crisis rating model is constructed based on the associated crisis rating heat map, and data training and verification are performed on the clustering results in the crisis rating model to obtain a mental health crisis rating model.

[0085] It should be noted that after obtaining the fluctuation state matrix expressing psychological activities described by metadata, it can be used to build an evaluation model. However, due to the large number of psychological activity fluctuation states in the fluctuation state matrix, it is difficult to determine the crisis rating corresponding to each psychological activity fluctuation state if it is directly imported into the model training, so it is difficult to know whether the psychological activity fluctuation state is a crisis or not. Therefore, it is necessary to correspond each psychological activity fluctuation state to each associated crisis rating one by one. This method obtains the color index of each associated crisis rating by coloring the associated crisis rating in the form of a heat map, and provides a cluster creation basis for the clustering of each psychological activity fluctuation state, so that the clustering of psychological activity fluctuation states and associated crisis ratings is more accurately corresponded. Then, the psychological activity state-color coding knowledge graph identifies the fluctuation state matrix to obtain the color coding sequence of each psychological activity fluctuation state in the matrix, calculates the current membership matrix of each sub-cluster cluster based on the color coding sequence, iteratively updates the sub-cluster cluster through the current membership matrix, and finally generates the clustering result, so as to realize clustering the color coding sequence of each psychological activity fluctuation state one by one according to the color of the associated crisis rating, which makes it possible to reasonably assign an accurate associated crisis rating to each psychological activity fluctuation state, effectively improve the evaluation performance of the rating model and the credibility of the warning results, and finally train the clustering result to construct a crisis rating model for evaluating students' mental health. The trained mental health crisis rating model is a display of the overall evaluation results of students' psychological activity changes.

[0086] It should be noted that this method can give a highly corresponding correlation crisis rating to each cluster of psychological activity fluctuation states in the psychological activity fluctuation state matrix described by the collected metadata, thereby making the construction of the mental health crisis rating model more accurate and reasonable, greatly improving the reliability of students' mental health assessment results, and reducing assessment errors. At the same time, it is helpful for the in-depth understanding and effective management of students' mental health, and will also provide valuable reference information for education policy makers. These monitoring and evaluation information will help to more accurately grasp the development direction of secondary vocational education, formulate education policies more scientifically, and better promote the comprehensive and healthy development of secondary vocational education.

[0087] More specifically, the step S108 includes the following steps:

[0088] Obtain the preset expected psychological rating range when the student's psychology is in a healthy standard state, and simultaneously obtain one or more benchmark psychological rating indicators corresponding to the expected psychological rating range for each structured mental health activity;

[0089] Obtaining the established qualified rating scores of one or more benchmark psychological rating indicators according to the expected psychological rating range, and comprehensively estimating the qualified rating standard score threshold of each structured mental health activity through the established qualified rating scores;

[0090] Constructing a qualified rating model, fitting points in the qualified rating model based on the qualified rating standard score thresholds of each structured mental health activity, and obtaining a standardized mental health rating model for students' mental health;

[0091] Creating a rating distribution network of structured mental health activities, extracting the rating distribution network in the mental health crisis rating model, defined as a first rating distribution network, and extracting the rating distribution network of the standardized mental health rating model, defined as a second rating distribution network;

[0092] Preset a mapping cost function based on the expected psychological rating range, map the first rating distribution network to the second rating distribution network with the mapping cost function as the mapping purpose, and obtain an actual joint distribution set of rating distributions;

[0093] Preset a warning threshold when the student's mental health reaches a dangerous rating, establish an abnormal joint distribution set based on the warning threshold, and introduce the Sinkhorn distance method to calculate the distribution offset distance between the actual joint distribution set and the abnormal joint distribution set;

[0094] If the distribution offset distance is greater than the preset distribution offset distance, the student is marked as a normal mental health student and no warning signal is generated;

[0095] If the distribution offset distance is less than the preset distribution offset distance, the student is marked as an abnormally mentally healthy student. At this time, a warning signal with a corresponding rating is generated based on the distribution offset distance to issue a mental health warning.

[0096] It should be noted that when the mental health crisis rating model of students participating in different structured activities is obtained, it is equivalent to knowing the student's mental health assessment score. However, the assessment score is presented in the form of a model, and the model contains the assessment results of different psychological rating indicators. If the preset score comparison method or the similarity comparison of the data model is used directly, it may lead to a large deviation in the final output rating analysis results, resulting in early warning errors in mental health monitoring. In this regard, this method estimates the qualified rating standard score threshold of students participating in each structured mental health activity in a healthy standard state by using the benchmark psychological rating indicators rated by the mental health crisis rating model, and then further constructs a standardized mental health crisis rating model based on these qualified rating standard score thresholds. Since one of the two is a standardized model and the other is a real-time model constructed by real-time monitoring of students' psychological activities, there will be differences in the rating distribution of each structured mental health activity between the two. Therefore, this method maps the first rating distribution network of the mental health crisis rating model to the second rating distribution network of the standardized mental health rating model. At this time, the two rating distribution networks after the mapping will reflect the distribution difference of the data network model, that is, the distribution offset distance between the actual joint distribution set and the abnormal joint distribution set. If the distribution offset distance is greater than the preset distribution offset distance, it means that the rating of the change of the student's psychological activity state does not meet the warning standard, that is, the psychological activity state is healthy, so it is marked as a normal mental health student and no warning signal is generated; otherwise, it means that the rating of the change of the student's psychological activity state has reached the warning threshold, and the psychological activity state is unhealthy, so it is marked as an abnormal mental health student and a warning signal is generated for warning. This method can reliably analyze the differences in rating distribution between the mental health rating model and the standardized model that display students' mental health assessment scores, thereby making the early warning results based on the mental health rating more accurate, ensuring that early warning signals can be issued in a timely manner when students' mental conditions fluctuate or become abnormal, so that corresponding intervention measures can be taken to improve the intelligence of mental health early warning.

[0097] It should be noted that the Chinese name of the Sinkhorn distance method is the entropy regularized distance method, which is used to measure the difference between two probability distributions. The Sinkhorn distance method optimizes computational efficiency and smoothness, overcomes the problem of high cost of distribution distance calculation, and further improves the accuracy of distribution shift between the actual joint distribution set and the abnormal joint distribution set.

[0098] In addition, the AI-enabled student mental health early warning method further includes the following steps:

[0099] Obtain the current mental health warning results of the target students, and at the same time obtain the intervention plan formulated by the student mental health warning system for the current mental health warning results;

[0100] Intervene the current mental health warning result of the target student based on the intervention counseling program, reacquire the mental health warning result after the intervention, calculate the cosine similarity between the mental health warning result after the intervention and the current mental health warning result, and determine the one-time counseling recovery rate of the intervention program according to the cosine similarity;

[0101] Determine whether the one-time counseling recovery rate is less than the preset counseling recovery rate. If so, obtain the intervention time planning information of the intervention plan, and extract the intervention time node array through the intervention time planning information;

[0102] Preset an ideal drainage recovery rate range, rearrange the intervention time node array, define the ideal drainage recovery rate range as the objective function, and calculate the gradient vector of the objective function between each intervention time node in the intervention time node array;

[0103] Iteratively update each time node in the intervention time node array based on the gradient vector until the update stops after a preset iteration frequency, and finally output a new intervention time node array;

[0104] Based on the new intervention time node array, the intervention plan is optimized and secondary intervention is performed on the target students. If the counseling recovery rate still does not reach the preset counseling recovery rate, the local weighted regression algorithm is introduced to calculate the intervention benefit regression coefficient between the fitted secondary counseling recovery rate and the new intervention time node array. The intervention intensity for the current students is optimized based on the intervention benefit regression coefficient.

[0105] It should be noted that when students' mental health is abnormal, the mental health early warning system will formulate targeted intervention plans for students to relieve their mental illness problems. However, the intervention time planning of the intervention plan may be uneven or irregular, which will lead to inadequate intervention or a long time interval when the intervention plan is used to relieve students' psychology, causing students to become lazy and slack, which greatly reduces the quality and effectiveness of the intervention plan. Therefore, this method can further re-plan the intervention time of the intervention plan based on whether the rehabilitation rate after the intervention plan reaches the standard, with the ideal rehabilitation rate range as the target, so that the intervention plan for target students with mental health problems is more reasonable and efficient in the time dimension, improves the counseling quality of the intervention plan, and prevents the target students from becoming lazy and slack due to insufficient or too long intervention time, with high reliability.

[0106] In addition, the secondary intervention is performed on the target students based on the new intervention time node array. If the counseling recovery rate still does not reach the preset counseling recovery rate, a local weighted regression algorithm is introduced to calculate the intervention benefit regression coefficient between the fitted secondary counseling recovery rate and the new intervention time node array. The intervention intensity for the current students is optimized according to the intervention benefit regression coefficient, which specifically includes the following steps:

[0107] The intervention plan is optimized according to the new intervention time node array, and the target students are re-intervened to obtain the secondary counseling recovery rate. If the secondary counseling recovery rate still does not reach the preset counseling recovery rate, a local weighted regression algorithm is introduced, and the Gaussian kernel function and bandwidth base are preset in the local weighted regression algorithm;

[0108] Based on the Gaussian kernel function, the weight value between the secondary counseling recovery rate of the target student and each intervention time node in the new intervention time node array is calculated to obtain multiple intervention benefit weight functions, and an intervention benefit weight matrix is ​​constructed through the multiple weight functions;

[0109] Perform regression fitting according to the intervention benefit weight matrix until it meets the bandwidth base, obtain a local weighted regression model of the intervention benefit, solve the local weighted regression model, and obtain the intervention benefit regression coefficient;

[0110] The intervention benefit regression coefficient obtained through the local weighted regression model is used to determine the intervention intensity for the current students. Based on the intervention intensity, the intervention content is increased and the intervention strategy is adjusted to optimize the guidance effect of the intervention plan.

[0111] It should be noted that after the intervention time of the intervention plan is replanned, if the counseling recovery rate still does not meet the standard, then at this time, if there is no problem with the intervention time planning, the intervention strategy of the intervention plan should be optimized. When the intervention strategy and intervention content of the intervention plan are relatively dry or the benefits are low, the target students will also have low intervention quality, which will aggravate the psychological problems of the target students. In view of this, this method uses the local weighted regression method to calculate and analyze the regression coefficient between the secondary counseling recovery rate and each intervention time node in the new intervention time node array, so as to know the intervention benefit of the optimized intervention plan, increase the intervention content in the intervention plan and adjust the intervention strategy according to the intervention benefit, so as to achieve more efficient and in-depth mental health intervention, make the counseling effect of the intervention plan more accurate and considerable, and reduce the recurrence rate of students' mental health problems or the probability of complications.

[0112] The second aspect of the present invention provides a student mental health early warning system based on AI empowerment, such as Figure 3As shown, the student mental health warning system includes a memory 41 and a processor 42. The memory 41 stores a student mental health warning method program based on AI empowerment. When the student mental health warning method program is executed by the processor 42, any one of the steps of the student mental health warning method is implemented.

[0113] The third aspect of the present invention provides a computer-readable storage medium, which stores a computer detection program. When the computer detection program is executed by at least one processor, it can implement any one of the AI-enabled student mental health early warning methods.

[0114] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A student mental health early warning method based on AI empowerment, characterized in that: The following steps are involved: S102: Obtain several structured mental health activities of students' mental health and their event variable rules and necessary triggering behaviors, construct event triggers and event listeners according to the event variable rules and necessary triggering behaviors, collect students' mental health activity data in real time through event listeners and event triggers, and obtain metadata; S104: Obtaining the psychological behavior norms of the students' mental health, constructing a psychological activity state differential equation when the students perform structured mental health activities based on the psychological behavior norms, solving the psychological activity state differential equation with metadata, and obtaining a fluctuation state matrix of the students' psychological activities when each metadata is generated; S106: drawing a correlation crisis rating heat map based on the correlation crisis rating of the students' mental health, clustering each mental activity fluctuation state of the fluctuation state matrix into the correlation crisis rating heat map, obtaining a clustering result, and using the clustering result to train a rating model to generate a mental health crisis rating model; S108: A standardized mental health rating model is fitted through the expected psychological rating range preset when the students are in a healthy standard state, the mental health crisis rating model is mapped to the standardized mental health rating model and the distribution distance deviation is analyzed to provide early warning for students with abnormal mental health.

2. According to claim 1, a student mental health early warning method based on AI empowerment is characterized in that: The step S102 specifically includes the following steps: Obtaining preset monitoring requirements for students' mental health, and extracting several structured mental health activities required for monitoring to assess students' mental health through the preset monitoring requirements, which are defined as susceptible structured events; Obtaining event variable rules for each susceptible structured event, and setting the event type of each susceptible structured event and the necessary triggering behavior corresponding to each event type according to the event variable rules; Create an event trigger architecture based on the event type of each susceptible structured event, set the event source according to the necessary trigger behavior corresponding to each event type, and follow the event variable rules to generate the event source's capture decision for each event type; Using the capture decision of each event type by the event source, an event trigger architecture is written to obtain an event trigger for each susceptible structured event with necessary triggering behavior; Obtain the historical operation logic of the event type corresponding to the necessary triggering behavior operation after the event trigger executes the capture decision, obtain the trigger response function of the necessary triggering behavior in the big data based on the historical operation logic, and build a series of event listeners that can be associated with the event type according to the trigger response function; Stripping the system node network of each susceptible structured event, deploying a series of event listeners in the system node network corresponding to the susceptible structured event, and monitoring the real-time event source of each necessary triggering behavior operation corresponding to the event type in the system node network through event triggers to obtain the real-time trigger result of each event trigger; When a series of event listeners receive the corresponding real-time trigger results, the hash state functions between the real-time event source and different historical operation logics are calculated one by one, and only the historical operation logic corresponding to the maximum hash state function is extracted and marked as the established operation logic; The event listener that receives the real-time trigger result is controlled to execute the established operation logic to obtain multiple metadata of the student's structured mental health activities at the current timestamp.

3. According to the AI-enabled student mental health early warning method of claim 1, it is characterized in that: The step S104 specifically includes the following steps: Obtain the psychological behavior norms of students' mental health, and use the psychological behavior norms as the constraint benchmark to obtain the psychological activity flow when structured mental health activities are executed according to event variable rules under the constraint benchmark; Introducing the difference method to obtain the differential linear invariant function of the psychological activity process, if the differential linear invariant function is within a preset differential linear invariant function interval, then calculating and constructing the activity state of the psychological activity process in the difference method based on the differential linear invariant function to generate a psychological activity state differential equation; The expected state space time series of the current timestamp under the preset differential linear invariant function is used to convert the multiple metadata of the student's structured mental health activities at the current timestamp into state variables at the current timestamp time series based on the expected state space time series, which are defined as mental activity state variables; Derivation of a differential relationship of metadata expressing psychological activity fluctuations based on psychological activity state variables, substituting the psychological activity state variables into the psychological activity state differential equation for updating, and generating a derived and updated psychological activity state variable; Substituting the updated mental activity state variables derived from the derivation back into the mental activity state difference equation to perform reconstruction processing, thereby obtaining a new mental activity state difference equation; Solve the differential equation of the psychological activity state of the freshmen to obtain the state space expression of the change of the psychological activity state when each metadata is generated, and determine the fluctuation state matrix of the student's psychological activity when each metadata is generated during the psychological activity process according to the description of the state space expression.

4. According to the AI-enabled student mental health early warning method of claim 1, it is characterized in that: The step S106 specifically includes the following steps: Acquire crisis prevention standards and regulations for students' mental health, and acquire multiple related crisis ratings of students' mental health activities with different structured mental health activities based on the crisis prevention standards and regulations; Construct a heat map, retrieve the mental health assessment scores of students when their mental health reaches each associated crisis rating based on the mental health management log, fit the mental health assessment scores corresponding to each associated crisis level in the heat map, and generate an associated crisis rating heat map; Obtain the color index of each associated crisis rating through the associated crisis rating heat map, obtain the psychological activity state-color coding knowledge map based on the big data network, use the psychological activity state-color coding knowledge map to identify the fluctuation state matrix, and obtain the color coding sequence of each psychological activity fluctuation state; Constructing N sub-cluster clusters based on the color index of each associated crisis rating, calculating the membership of each sub-cluster cluster through the color coding sequence of each psychological activity fluctuation state, and obtaining a number of current membership matrices; Calculating the center vector of each sub-cluster according to each current membership matrix to obtain multiple new sub-cluster centers, iteratively updating the membership of each mental activity fluctuation state in the fluctuation state matrix according to the multiple new sub-cluster centers, and outputting the distance offset value of each membership matrix; A maximum distance offset threshold is preset, and the above-mentioned update iterative steps are repeated to continuously calculate the distance offset value of the membership matrix. When the distance offset value is less than the maximum distance offset threshold, the iterative update is stopped, and finally a clustering result of each psychological activity fluctuation state in the fluctuation state matrix is ​​generated, which is clustered in each associated crisis rating; A crisis rating model is constructed based on the associated crisis rating heat map, and data training and verification are performed on the clustering results in the crisis rating model to obtain a mental health crisis rating model.

5. According to the AI-enabled student mental health early warning method of claim 1, it is characterized in that: The step S108 specifically includes the following steps: Obtain the preset expected psychological rating range when the student's psychology is in a healthy standard state, and simultaneously obtain one or more benchmark psychological rating indicators corresponding to the expected psychological rating range for each structured mental health activity; Obtaining the established qualified rating scores of one or more benchmark psychological rating indicators according to the expected psychological rating range, and comprehensively estimating the qualified rating standard score threshold of each structured mental health activity through the established qualified rating scores; Constructing a qualified rating model, fitting points in the qualified rating model based on the qualified rating standard score thresholds of each structured mental health activity, and obtaining a standardized mental health rating model for students' mental health; Creating a rating distribution network of structured mental health activities, extracting the rating distribution network in the mental health crisis rating model, defined as a first rating distribution network, and extracting the rating distribution network of the standardized mental health rating model, defined as a second rating distribution network; Preset a mapping cost function based on the expected psychological rating range, map the first rating distribution network to the second rating distribution network with the mapping cost function as the mapping purpose, and obtain an actual joint distribution set of rating distributions; Preset a warning threshold when the student's mental health reaches a dangerous rating, establish an abnormal joint distribution set based on the warning threshold, and introduce the Sinkhorn distance method to calculate the distribution offset distance between the actual joint distribution set and the abnormal joint distribution set; If the distribution offset distance is greater than the preset distribution offset distance, the student is marked as a normal mental health student and no warning signal is generated; If the distribution offset distance is less than the preset distribution offset distance, the student is marked as an abnormally mentally healthy student. At this time, a warning signal with a corresponding rating is generated based on the distribution offset distance to issue a mental health warning.

6. A student mental health early warning system based on AI empowerment, characterized by: The student mental health early warning system includes a memory and a processor. The memory stores a student mental health early warning method program based on AI empowerment. When the student mental health early warning method program is executed by the processor, the student mental health early warning method steps as described in any one of claims 1-5 are implemented.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer detection program, and when the computer detection program is executed by at least one processor, it implements the student mental health early warning method based on AI empowerment as described in any one of claims 1-5.

Images