Deep prediction and intervention method for college students' social anxiety tendency based on fusion of behavioral data

Abstract

This invention belongs to the field of mental health prediction and intervention technology. It discloses a method for in-depth prediction and intervention of social anxiety tendencies in college students using integrated behavioral data. The method includes: acquiring integrated behavioral data of target college students; constructing a multi-level association network based on the integrated behavioral data; determining whether the target college students have social anxiety tendencies; conducting in-depth analysis of the integrated behavioral data of target college students with social anxiety tendencies; determining the social anxiety level of the target college students and generating a risk profile; matching the social anxiety level and risk profile with a preset multi-level intervention response framework to determine a set of candidate intervention strategies; selecting from the candidate intervention strategy set based on reward and punishment signals; determining the optimal intervention plan; analyzing feedback data to determine feedback signals; and dynamically adjusting the selection logic of the optimal intervention plan. This method forms a closed loop through prediction-evaluation-intervention-optimization, improving the scientificity and accuracy of prediction and intervention.

Description

Technical Field

[0001] This invention relates to the field of mental health prediction and intervention technology, and more specifically, to a method for in-depth prediction and intervention of social anxiety tendencies among college students by integrating behavioral data. Background Technology

[0002] Social anxiety is a common psychological problem among college students, which seriously affects their academic development and interpersonal communication. Timely and accurate assessment of the depth of their social anxiety tendencies and effective intervention are crucial for protecting their mental health and promoting their social functioning.

[0003] In existing technologies, some universities introduce behavioral data to identify social anxiety. They establish risk identification models based on factors such as internet usage frequency, social media interactions, or activity trajectories. Then, they perform statistical and static analysis on the data at a certain point in time or over a certain period of time to output a single anxiety risk level. However, social anxiety among college students is a highly dynamic psychological state that fluctuates with context and time. Static analysis is unable to capture such instantaneous changes and long-term trends. For example, if a user stays in the dormitory for a long time, the static model cannot distinguish whether this is due to avoidance behavior caused by social anxiety or normal study or rest, which leads to errors in risk identification and failure of intervention strategies.

[0004] In view of this, the present invention proposes a method for predicting and intervening in the social anxiety tendency of college students by integrating behavioral data in order to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the existing technology and to achieve the above objectives, the present invention provides the following technical solution: a method for in-depth prediction and intervention of social anxiety tendencies among college students by integrating behavioral data, comprising: S1: Obtain integration behavior data of target college students, construct a multi-level association network based on the integration behavior data, and determine whether the target college students have a tendency of social anxiety; S2: If it is determined that the target college student has a tendency of social anxiety, then conduct in-depth analysis of the target college student's integration behavior data, determine the level of social anxiety of the target college student and generate a risk profile; S3: Match the social anxiety level and risk profile with the pre-set multi-level intervention response framework to determine the candidate intervention strategy set. Select the candidate intervention strategy set based on reward and punishment signals to determine the optimal intervention plan. S4: Obtain feedback data after implementing the optimal intervention plan, analyze the feedback data to determine the feedback signal, and dynamically adjust the selection logic of the optimal intervention plan.

[0006] Furthermore, methods for constructing multi-level interconnected networks include: A multi-level association network is constructed, which includes a primary central judgment node for outputting comprehensive tendency indicators, a secondary primary node for aggregating tendency contributions to obtain key behavioral factor quantities, and a tertiary secondary node for receiving fused behavioral data. The fused behavioral data is input into the corresponding tertiary secondary node. Information transmission rules and weights between corresponding nodes in adjacent levels are preset. The corresponding nodes in adjacent levels are connected according to the information transmission rules and weights to complete the construction of the multi-level association network.

[0007] Furthermore, methods for determining whether target college students exhibit social anxiety tendencies include: The pre-defined mapping rules convert the fusion behavior data in each third-level secondary node into the corresponding tendency contribution. Based on the weight between the tendency contribution in each third-level secondary node and the corresponding second-level primary node, the key behavior factor quantity in each second-level primary node is determined. Based on the weight between the key behavior factor quantity in each second-level primary node and the first-level central judgment node, the comprehensive tendency index of the target college students is determined. A preset threshold is set. When the comprehensive tendency index is greater than the threshold, the target college student is judged to have a tendency towards social anxiety; otherwise, they are judged not to have a tendency towards social anxiety.

[0008] Furthermore, methods for integrating behavioral data for in-depth analysis include: Based on a preset time window mechanism, the fused behavioral data is divided into perceptual structure segments. The perceptual structure segments are then factored to obtain the objective behavioral factors and subjective perceptual factors corresponding to each segment, and arranged in chronological order as objective behavioral sequences and subjective perceptual sequences. A dynamic sequence prediction model is constructed, which takes the objective behavioral sequence as the model input, the subjective perception sequence as the supervision signal, and outputs a social anxiety prediction score sequence. By determining the corresponding behavior-perception vector for each objective behavior factor and subjective perception factor, and determining the evolutionary pattern factor based on the evolutionary characteristics of the behavior-perception vector, the changing trend of each evolutionary pattern factor is marked as the pattern label of each behavior-perception vector, and the behavior-perception vectors are arranged to form a sequence of behavior-psychological correlation evolutionary patterns.

[0009] Furthermore, methods for determining the level of social anxiety among target university students include: The average level, fluctuation range and jump frequency indicators are obtained from the social anxiety prediction score sequence. The pattern change frequency indicator is determined based on the number of pattern label switching in the behavior-psychology association evolution pattern sequence. Low, medium, and high threshold ranges are preset for each indicator. The threshold range in which each indicator falls is determined. Based on the worst-case risk assessment mechanism, a comprehensive judgment is made on the threshold ranges of each indicator to determine the social risk level of the target university student. The judgment method is as follows: If any indicator is in the high threshold range, the target college student is judged to have severe anxiety; if no indicator is in the high threshold range, and at least two indicators are in the medium threshold range, the student is judged to have moderate anxiety; if all indicators are in the low risk range or only one indicator is in the medium threshold range, the student is judged to have low anxiety.

[0010] Furthermore, methods for generating risk profiles include: A double-helix risk network is constructed, which includes an input layer, a spiral channel layer, a coupling engine layer, and a label generation layer. The spiral channel layer includes the anxiety spiral channel and the behavior spiral channel. The anxiety spiral channel forms a time series graph of the social anxiety prediction score sequence received by the input layer and determines the fluctuation feature vector; the behavior spiral channel forms a behavioral psychological tensor of the behavioral-psychological association evolution pattern sequence received by the input layer and determines the behavioral cluster center trajectory. The coupling engine layer obtains the coupling feature set from the fluctuation feature vector and the behavior cluster center trajectory. The coupling feature set includes the mean of the behavior cluster center, the time-frequency joint feature and the synchronization jump component. The label generation layer outputs a triple risk label based on the coupled feature set. The triple risk label includes the dominant behavioral pattern, anxiety score fluctuation type, and jump sensitivity category label. A risk profile is formed based on the triple risk label.

[0011] Furthermore, methods for determining the set of candidate intervention strategies include: A multi-level intervention response framework is constructed, which divides response levels according to social risk level. Each response level contains several intervention programs, and each intervention program is associated with a set of adaptation condition vectors, including dominant behavioral patterns, anxiety score fluctuation types, and change sensitivity categories. The corresponding response levels are associated with the social risk level of the target college students, and the matching degree score is obtained by matching the adaptation condition vectors of each intervention program in the corresponding response level with the risk profile of the target college students. Set a matching threshold, filter out intervention programs with matching scores greater than the threshold, and prioritize them according to matching scores to determine the candidate intervention program set for the target college students.

[0012] Furthermore, methods for determining the optimal intervention plan include: Obtain the reward and punishment signals of each intervention plan in the candidate intervention plan set during the historical push process; dynamically adjust the priority of each intervention plan in the candidate intervention plan set based on the reward and punishment signals, determine the intervention plan with the highest priority as the optimal intervention plan, and implement the intervention on the target college students based on the optimal intervention plan.

[0013] Furthermore, methods for analyzing feedback data include: The study aims to obtain feedback data from target college students after implementing the optimal intervention plan. This data includes data on click behavior on intervention content, dwell time on intervention pages, and completion status of intervention guidance tasks. A quantitative score is used to determine the effectiveness of the feedback. An effective threshold is set, and feedback data with effectiveness scores above the threshold are designated as reward signals, while those below are designated as punishment signals.

[0014] Furthermore, methods for dynamically adjusting the selection logic of the optimal intervention plan include: If the reward / punishment signal corresponding to the intervention plan is a reward signal, the priority of the intervention plan in the candidate intervention strategy set will be increased by one level; conversely, if the reward / punishment signal is a punishment signal, the priority of the intervention plan in the candidate intervention strategy set will be decreased by one level.

[0015] The technical effects and advantages of this invention's method for in-depth prediction and intervention of social anxiety tendencies among college students by integrating behavioral data: 1. In order to overcome the difficulty of traditional analysis in comprehensively and objectively capturing the characteristics of social anxiety among college students, this invention constructs a multi-level association network, incorporates online and offline integrated behavioral data of college students, forms a complete chain from micro-behavioral factors to macro-tendency judgment, and combines preset mapping rules and weights to quantify behavioral data into tendency contribution and comprehensive indicators. In a multi-level semantic structure, it realizes the step-by-step identification and judgment of social anxiety tendencies, thereby capturing social anxiety characteristics more comprehensively and objectively, accurately judging whether college students have social anxiety tendencies, and improving the scientificity and accuracy of prediction. 2. To overcome the technical problem that static analysis cannot capture instantaneous changes, this invention obtains a social anxiety prediction score sequence and a behavioral-psychological evolution pattern sequence through in-depth analysis of fused behavioral data. This allows for the acquisition of multiple indicators, such as average level, fluctuation amplitude and jump frequency, and pattern transition frequency. Through joint evaluation of these multiple indicators, the invention can accurately classify an individual's anxiety level. Furthermore, it visually presents students' anxiety risk characteristics. Using a double-helix risk network structure, it extracts a coupled feature set of anxiety temporal features and behavioral clustering features, outputting a triple risk label to form an individual risk profile. Combined with a multi-level intervention response framework, it intelligently matches a set of candidate intervention strategies based on the risk profile and adaptation condition vectors. Finally, it selects the optimal intervention plan based on reward and punishment signals, ensuring that intervention measures are tailored to individual student circumstances and enhancing the pertinence and effectiveness of the intervention. 3. This invention overcomes the problem of traditional interventions lacking a closed-loop feedback mechanism and making it difficult to dynamically adjust strategies based on students' actual responses to intervention measures. By acquiring feedback data after implementing the intervention plan, quantifying and determining the feedback effectiveness score, and differentiating reward and punishment signals based on the feedback effectiveness score, the optimal intervention plan selection logic is dynamically adjusted. Through continuous closed-loop feedback, subsequent intervention strategies are optimized, and intervention strategies are promptly optimized based on students' actual responses. This ensures that the intervention plan continuously meets students' needs, constantly improves the intervention effect, and forms a long-term mental health protection mechanism. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the structural method for predicting and intervening in the social anxiety tendency of college students by integrating behavioral data, as described in this invention. Figure 2 This is a flowchart illustrating the system for predicting and intervening in the social anxiety tendencies of college students by integrating behavioral data, as described in this invention. Figure 3 This is a schematic diagram illustrating the risk profile generated by the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1

[0018] Please see Figure 1 and Figure 3 As shown in this embodiment, the method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data has the following main design contents: Social anxiety is a common psychological problem among college students. Its manifestations are characterized by significant behavioral features and situational fluctuations. Traditional anxiety identification methods often rely on static questionnaires or single behavioral indicators. In real-world scenarios, college students' social anxiety is often the result of the long-term coupling and evolution of multiple behavioral factors. There are non-linear correlations between different behavioral data. For example, a reduction in the range of spatial activities may be related to anxiety avoidance behavior, but it may also stem from self-study arrangements during the period leading up to exams. It is difficult to make accurate distinctions based on a single indicator. Therefore, there is an urgent need for a judgment mechanism with multi-dimensional integration capabilities and structured expression capabilities. Based on this, a method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data was designed, including: S1: Obtain integration behavior data of target college students, construct a multi-level association network based on the integration behavior data, and determine whether the target college students have a tendency of social anxiety; Methods for constructing multi-level interconnected networks include: A multi-level association network is constructed, which includes a primary central judgment node for outputting comprehensive tendency indicators, a secondary primary node for aggregating tendency contributions to obtain key behavioral factor quantities, and a tertiary secondary node for receiving fused behavioral data. The fused behavioral data is input into the corresponding tertiary secondary node. Information transmission rules and weights between corresponding nodes in adjacent levels are preset. The corresponding nodes in adjacent levels are connected according to the information transmission rules and weights to complete the construction of the multi-level association network.

[0019] It should be explained that the multi-level association network is a system specifically constructed by this invention to accurately determine the social anxiety tendencies of college students, based on the hierarchical features and semantic associations of online and offline integrated behavioral data.

[0020] It should be explained that each type of fused behavioral data corresponds to a third-level secondary node. A second-level primary node connects several independent third-level secondary nodes through weights and information transmission rules. A first-level central decision node connects several independent second-level primary nodes through weights and information transmission rules. Information transmission rules are used to define the flow and transformation mechanism of data between nodes at different levels. For example, a certain second-level primary node only receives specific types of data from third-level secondary nodes. Weights are used to represent the degree of influence when information is transmitted from lower-level nodes to upper-level nodes. The weights between nodes depend on statistical analysis settings.

[0021] Methods for determining whether a target college student has a tendency toward social anxiety include: The pre-defined mapping rules convert the fusion behavior data in each third-level secondary node into the corresponding tendency contribution. Based on the weight between the tendency contribution in each third-level secondary node and the corresponding second-level primary node, the key behavior factor quantity in each second-level primary node is determined. Based on the weight between the key behavior factor quantity in each second-level primary node and the first-level central judgment node, the comprehensive tendency index of the target college students is determined. A preset threshold is set. When the comprehensive tendency index is greater than the threshold, the target college student is judged to have a tendency towards social anxiety; otherwise, they are judged not to have a tendency towards social anxiety.

[0022] It should be explained that the judgment threshold is determined through statistical analysis of a large amount of sample data. For example, the 85th percentile of the comprehensive tendency index can be set as the judgment threshold.

[0023] It should be explained that the integrated behavioral data includes, but is not limited to: daily average social interaction frequency, indoor and outdoor activity ratio, active time on online social platforms, classroom interaction frequency, and dormitory stay time. Each behavioral data is mapped according to its semantic attributes and historical statistical thresholds. For example, an interval mapping function is used to convert each behavioral data into a tendency contribution in the range of [-1, 1]. A positive value indicates an increased anxiety tendency, and a negative value indicates a decreased anxiety tendency. Based on the node connection weights between each secondary node and its corresponding primary node, the tendency contribution is weighted and summed to obtain the key behavioral factor quantity for each primary node. The key behavioral factor quantity includes, but is not limited to, social avoidance factor, attention fluctuation factor, and isolation factor. For example, the quantitative value of the avoidance factor can be obtained by weighting and summing the self-social frequency reduction, social platform deactivation, and spatial closure behaviors input by multiple secondary nodes with their corresponding weights.

[0024] In this embodiment, by constructing a multi-level association network including three-level secondary nodes, two-level primary nodes, and a first-level central judgment node, the original behavioral data can be effectively transformed into semantically clear key behavioral factors, which are then further mapped into comprehensive tendency indicators. This achieves layer-by-layer abstraction and feature aggregation of complex behavioral data into social anxiety tendency indicators, and timely response to subtle changes in fused behavioral data. As a result, the dynamic evolution of social anxiety can be captured, making anxiety tendency determination more scientific, systematic, and interpretable. Example 2

[0025] Please see Figure 2 As shown in this embodiment, the method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data has the following main design contents: Social anxiety among college students is not a static state, but an evolving process dynamically influenced by multiple factors such as environment, behavior, and subjective feelings. When conducting social anxiety assessments, some universities typically use the following methods: based on data from a single psychological questionnaire or regular interviews, or by regularly sampling and statically statistically analyzing behavioral data such as internet usage and activity patterns, ultimately outputting a fixed anxiety risk level or state label. However, during exam weeks, students experience reduced social activities and intense mood swings. At this time, the static model may misjudge as a persistent high risk, failing to reveal the evolutionary trend of anxiety, sensitive fluctuation points, dominant behavioral patterns, and other deep psychological mechanisms. It is also difficult to characterize the semantic differences of the same behavior in different individuals, at different times, and under different psychological backgrounds, resulting in poor generalization ability.

[0026] Based on this, a method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data was designed, including: S2: If it is determined that the target college student has a tendency of social anxiety, then conduct in-depth analysis of the target college student's integration behavior data, determine the level of social anxiety of the target college student and generate a risk profile; Methods for deep analysis by integrating behavioral data include: Based on a preset time window mechanism, the fused behavioral data is divided into perceptual structure segments. The perceptual structure segments are then factored to obtain the objective behavioral factors and subjective perceptual factors corresponding to each segment, and arranged in chronological order as objective behavioral sequences and subjective perceptual sequences. A dynamic sequence prediction model is constructed, which takes the objective behavioral sequence as the model input, the subjective perception sequence as the supervision signal, and outputs a social anxiety prediction score sequence. By determining the corresponding behavior-perception vector for each objective behavior factor and subjective perception factor, and determining the evolutionary pattern factor based on the evolutionary characteristics of the behavior-perception vector, the changing trend of each evolutionary pattern factor is marked as the pattern label of each behavior-perception vector, and the behavior-perception vectors are arranged to form a sequence of behavior-psychological correlation evolutionary patterns.

[0027] It should be explained that the preset time window mechanism can be flexibly configured according to the actual sampling frequency, dividing continuous fusion behavior data into independent segments. For example, one day can be set as a time window, dividing one month's fusion behavior data into 30 perceptual structure segments. The acquired objective behavioral factors are characteristic variables that can objectively reflect college students' social behavior, such as average movement distance and number of social interactions. The acquired subjective perceptual factors are characteristic variables of college students' subjective feelings and evaluations of social situations, such as self-rated anxiety and semantic labeling. The objective behavioral sequence shows the changes in college students' objective social behavior in different time windows; the subjective perceptual sequence reflects the evolution of college students' subjective psychological feelings in the same period.

[0028] It needs to be explained that, for each perceptual structure segment, its objective behavioral factors (such as travel frequency, activity range, social density, etc.) and subjective perceptual factors (such as anxiety scores, self-evaluation, subjective stress scale results, etc.) are extracted and concatenated to form a behavior-perceptual vector. Evolutionary pattern factors are used to identify the core feature type of behavioral and psychological linkage evolution in each time window (such as anxiety surge type, behavior closure type, and fluctuation relief type, etc.), identify its switching frequency, duration and mutation point, and convert them into pattern labels. All behavior-perceptual vector sequences with pattern labels are integrated to construct a behavior-psychological correlation evolutionary pattern sequence.

[0029] Methods for determining the social anxiety level of target college students include: The average level, fluctuation range and jump frequency indicators are obtained from the social anxiety prediction score sequence. The pattern change frequency indicator is determined based on the number of pattern label switching in the behavior-psychology association evolution pattern sequence. Low, medium, and high threshold ranges are preset for each indicator. The threshold range in which each indicator falls is determined. Based on the worst-case risk assessment mechanism, a comprehensive judgment is made on the threshold ranges of each indicator to determine the social risk level of the target university student. The judgment method is as follows: If any indicator is in the high threshold range, the target college student is judged to have severe anxiety; if no indicator is in the high threshold range, and at least two indicators are in the medium threshold range, the student is judged to have moderate anxiety; if all indicators are in the low risk range or only one indicator is in the medium threshold range, the student is judged to have low anxiety.

[0030] It should be explained that the low, medium, and high threshold ranges for each indicator were determined using an interval statistical method. It should be explained that the average level index is used to characterize the overall anxiety intensity of the target college students over a period of time; the fluctuation index is used to measure the degree of fluctuation in anxiety scores over time. The greater the fluctuation, the worse the emotional stability, the more sensitive the individual is to the stress response to external stimuli, and the higher the potential risk; the jump frequency index is used to assess the frequency of significant jumps or drops in anxiety scores, reflecting the abruptness of an individual's anxiety state and emotional control ability; the pattern change frequency index is used to determine the frequency of an individual rapidly switching from one behavioral and psychological pattern to another. High frequency of pattern switching usually means that the individual's behavioral regulation ability is unstable or emotional response is discontinuous.

[0031] Methods for generating risk profiles include: A double-helix risk network is constructed, which includes an input layer, a spiral channel layer, a coupling engine layer, and a label generation layer. The spiral channel layer includes the anxiety spiral channel and the behavior spiral channel. The anxiety spiral channel forms a time series graph of the social anxiety prediction score sequence received by the input layer and determines the fluctuation feature vector; the behavior spiral channel forms a behavioral psychological tensor of the behavioral-psychological association evolution pattern sequence received by the input layer and determines the behavioral cluster center trajectory. The coupling engine layer obtains the coupling feature set from the fluctuation feature vector and the behavior cluster center trajectory. The coupling feature set includes the mean of the behavior cluster center, the time-frequency joint feature and the synchronization jump component. The label generation layer outputs a triple risk label based on the coupled feature set. The triple risk label includes the dominant behavioral pattern, anxiety score fluctuation type, and jump sensitivity category label. A risk profile is formed based on the triple risk label.

[0032] It should be explained that the double helix risk network is a feature specifically constructed by this invention to generate individualized risk profiles of social anxiety among college students, targeting the coupling relationship between the temporal characteristics of anxiety and the clustering characteristics of behavior.

[0033] It needs to be explained that the Anxiety Spiral Channel transforms the social anxiety score sequence into a time-series graph, explores the fluctuation patterns and transition trends of the social anxiety score sequence on the time axis, and extracts fluctuation feature vectors, which include peak values, amplitudes, and fluctuation frequencies. The Behavioral Spiral Channel performs tensor quantization on the behavioral-psychological correlation evolution pattern, and then extracts the cluster center trajectory of the tensor through a clustering algorithm to identify the college student's dominant behavioral preferences within a certain period.

[0034] The mean value of the cluster centers is obtained by averaging the behavior cluster center trajectories within a preset time window; the frequency distribution of the anxiety score sequence is extracted by performing wavelet transform, and then jointly encoded with the temporal change rate of the behavior trajectory to obtain time-frequency joint features, thereby capturing whether there is a high-frequency coupling phenomenon between behavior and anxiety; a "jump point" is defined, and the temporal matching degree of two jump point sequences is calculated to obtain the synchronous jump score, thereby assessing the individual's emotional sensitivity to event changes.

[0035] Based on the mean of behavioral cluster centers, the most representative behavioral patterns in the current stage are identified as the dominant behavioral pattern labels. Behavioral patterns include social isolation, information overload, and task avoidance. The synchronous jump score is compared with a preset threshold to determine whether the jump sensitivity category label is low, medium, or high. Based on the amplitude, frequency, waveform complexity, and time-frequency joint features in the fluctuation feature vector, the anxiety score fluctuation type is determined to be stable, highly volatile, or frequently jumping. In this embodiment, the perception segments are divided by a preset time window, objective and subjective factors are extracted, and an evolutionary pattern sequence is constructed based on the behavior-perception vector, thereby more realistically depicting the dynamic coupling relationship between college students' behavior and psychology. The social anxiety level identification is more granular and personalized by comprehensively judging through multi-dimensional indicators such as average level, fluctuation amplitude, jump frequency and pattern change frequency, which solves the technical bottleneck of single anxiety judgment dimension and poor sensitivity in the existing methods. A double-helix risk network is used to generate individualized risk profiles. By using dual-channel input of anxiety score time series and behavioral evolution trajectory, and combining the coupling engine to extract key coupling features, triple risk labels are extracted for dominant behavioral patterns, anxiety fluctuation types, and dyssensitivity. Based on this, a risk profile is constructed, which not only improves the ability to identify individual differences in social anxiety, but also provides a structured label basis for the subsequent implementation of precise intervention strategies. Example 3

[0036] Please see Figure 1-3As shown in this embodiment, the method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data has the following main design contents: Current interventions for social anxiety among college students often rely on fixed content delivery or appointment-based psychological counseling, which suffer from slow response, low strategy matching, and delayed feedback on intervention effects. For example, some platforms only push standardized intervention content based on the results of a single questionnaire, making it difficult to adapt to the different causes and evolutionary characteristics of anxiety among individuals. In addition, the lack of a dynamic optimization mechanism for the effectiveness of intervention strategies leads to the continued use of some intervention measures that are ineffective for a long time, reducing the overall intervention efficiency and resource utilization.

[0037] Based on this, a method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data was designed, including: S3: Match the social anxiety level and risk profile with the pre-set multi-level intervention response framework to determine the candidate intervention strategy set. Select the candidate intervention strategy set based on reward and punishment signals to determine the optimal intervention plan. S4: Obtain feedback data after implementing the optimal intervention plan, analyze the feedback data to determine the feedback signal, and dynamically adjust the selection logic of the optimal intervention plan.

[0038] Methods for determining the set of candidate intervention strategies include: A multi-level intervention response framework is constructed, which divides response levels according to social risk level. Each response level contains several intervention programs, and each intervention program is associated with a set of adaptation condition vectors, including dominant behavioral patterns, anxiety score fluctuation types, and change sensitivity categories. The corresponding response levels are associated with the social risk level of the target college students, and the matching degree score is obtained by matching the adaptation condition vectors of each intervention program in the corresponding response level with the risk profile of the target college students. Set a matching threshold, filter out intervention programs with matching scores greater than the threshold, and prioritize them according to matching scores to determine the candidate intervention program set for the target college students.

[0039] It should be explained that the similarity between the risk profile of the target college student and the fit condition vector of each intervention plan is calculated by using the cosine of the vector angle. This forms the matching score. The higher the matching score, the more suitable the intervention strategy is for the current student's anxiety characteristics.

[0040] It should be explained that the matching threshold was determined through retrospective analysis of intervention implementation records of a large number of historical individual samples.

[0041] Methods for determining the optimal intervention plan include: Obtain the reward and punishment signals of each intervention plan in the candidate intervention plan set during the historical push process; dynamically adjust the priority of each intervention plan in the candidate intervention plan set based on the reward and punishment signals, determine the intervention plan with the highest priority as the optimal intervention plan, and implement the intervention on the target college students based on the optimal intervention plan.

[0042] Methods for analyzing feedback data include: The study aims to obtain feedback data from target college students after implementing the optimal intervention plan. This data includes data on click behavior on intervention content, dwell time on intervention pages, and completion status of intervention guidance tasks. A quantitative score is used to determine the effectiveness of the feedback. An effective threshold is set, and feedback data with effectiveness scores above the threshold are designated as reward signals, while those below are designated as punishment signals.

[0043] It should be explained that the quantitative scoring rules for feedback data are as follows: if a user clicks on the intervention content, the click behavior is awarded 1 point, otherwise no point is awarded; if the user stays on the intervention page for more than or equal to 30 seconds, the stay behavior is awarded 1 point, otherwise no point is awarded; if the user completes the guided task set in the intervention content, the task completion behavior is awarded 1 point, otherwise no point is awarded. Based on the above three items, the calculated feedback effectiveness score ranges from 0 to 3 points. The preset effectiveness threshold range is 2 to 3 points. By comparing the effectiveness threshold range with the feedback effectiveness score, when the feedback effectiveness score is greater than or equal to the effectiveness threshold, it indicates that the current intervention has a good response effect on the target college student, and the intervention plan is given a reward signal. Conversely, it indicates that the intervention plan has a poor intervention response effect in the current individual, and its subsequent push priority should be reduced. Therefore, the intervention plan is given a penalty signal.

[0044] Methods for dynamically adjusting the selection logic of the optimal intervention plan include: If the reward / punishment signal corresponding to the intervention plan is a reward signal, the priority of the intervention plan in the candidate intervention strategy set will be increased by one level; conversely, if the reward / punishment signal is a punishment signal, the priority of the intervention plan in the candidate intervention strategy set will be decreased by one level.

[0045] In this embodiment, by introducing a multi-level intervention response framework and a feedback-driven intervention optimization mechanism, personalized matching and dynamic adaptive adjustment of intervention strategies are achieved. With the help of structured labels of social anxiety level and risk profile, candidate intervention plans that are highly consistent with individual risk characteristics can be screened from the corresponding response level, solving the problem of "one policy for a thousand people" in existing methods. At the same time, by setting matching thresholds and introducing historical reward and punishment signals, the priority of intervention plans is dynamically adjusted based on feedback data, and the intervention path is continuously optimized, providing a closed-loop support mechanism for intervention decision-making, thereby significantly improving the scientific nature, real-time performance and iterative optimization capability of intervention response. Example 4

[0046] Please see Figure 3 As shown in this embodiment, the system for in-depth prediction and intervention of college students' social anxiety tendencies by integrating behavioral data includes: The data acquisition module obtains data on the integration behavior of target college students; The multi-level network module constructs a network containing first-, second-, and third-level nodes, converts data into tendency contribution, and calculates a comprehensive tendency index to determine whether there is a tendency towards social anxiety. The in-depth analysis module divides data into segments, generates scoring and evolutionary pattern sequences, determines anxiety levels, and creates risk profiles with triple risk labels. The intervention execution module matches a multi-level framework to determine a set of candidate strategies and selects the optimal solution for implementation based on reward and punishment signals. The feedback optimization module analyzes feedback data to determine reward and punishment signals and dynamically adjusts strategy priorities.

[0047] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0048] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0049] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

[0050] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data, characterized in that... The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data includes: S1: Obtain integration behavior data of target college students, construct a multi-level association network based on the integration behavior data, and determine whether the target college students have a tendency of social anxiety; S2: If it is determined that the target college student has a tendency of social anxiety, then conduct in-depth analysis of the target college student's integration behavior data, determine the level of social anxiety of the target college student and generate a risk profile; S3: Match the social anxiety level and risk profile with the pre-set multi-level intervention response framework to determine the candidate intervention strategy set. Select the candidate intervention strategy set based on reward and punishment signals to determine the optimal intervention plan. S4: Obtain feedback data after implementing the optimal intervention plan, analyze the feedback data to determine the feedback signal, and dynamically adjust the selection logic of the optimal intervention plan.

2. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data as described in claim 1, characterized in that, The method for constructing a multi-level interconnected network includes: A multi-level association network is constructed, which includes a primary central judgment node for outputting comprehensive tendency indicators, a secondary primary node for aggregating tendency contributions to obtain key behavioral factor quantities, and a tertiary secondary node for receiving fused behavioral data. The fused behavioral data is input into the corresponding tertiary secondary node. Information transmission rules and weights between corresponding nodes in adjacent levels are preset. The corresponding nodes in adjacent levels are connected according to the information transmission rules and weights to complete the construction of the multi-level association network.

3. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data as described in claim 1, characterized in that, The methods for determining whether a target college student has a tendency towards social anxiety include: The pre-defined mapping rules convert the fusion behavior data in each third-level secondary node into the corresponding tendency contribution. Based on the weight between the tendency contribution in each third-level secondary node and the corresponding second-level primary node, the key behavior factor quantity in each second-level primary node is determined. Based on the weight between the key behavior factor quantity in each second-level primary node and the first-level central judgment node, the comprehensive tendency index of the target college students is determined. A preset threshold is set. When the comprehensive tendency index is greater than the threshold, the target college student is judged to have a tendency towards social anxiety; otherwise, they are judged not to have a tendency towards social anxiety.

4. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data according to claim 1, characterized in that, The methods for deep analysis of the fused behavioral data include: Based on a preset time window mechanism, the fused behavioral data is divided into perceptual structure segments. The perceptual structure segments are then factored to obtain the objective behavioral factors and subjective perceptual factors corresponding to each segment, and arranged in chronological order as objective behavioral sequences and subjective perceptual sequences. A dynamic sequence prediction model is constructed, which takes the objective behavioral sequence as the model input, the subjective perception sequence as the supervision signal, and outputs a social anxiety prediction score sequence. By determining the corresponding behavior-perception vector for each objective behavior factor and subjective perception factor, and determining the evolutionary pattern factor based on the evolutionary characteristics of the behavior-perception vector, the changing trend of each evolutionary pattern factor is marked as the pattern label of each behavior-perception vector, and the behavior-perception vectors are arranged to form a sequence of behavior-psychological correlation evolutionary patterns.

5. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data according to claim 1, characterized in that, The methods for determining the social anxiety level of target college students include: The average level, fluctuation range and jump frequency indicators are obtained from the social anxiety prediction score sequence. The pattern change frequency indicator is determined based on the number of pattern label switching in the behavior-psychology association evolution pattern sequence. Low, medium, and high threshold ranges are preset for each indicator. The threshold range in which each indicator falls is determined. Based on the worst-case risk assessment mechanism, a comprehensive judgment is made on the threshold ranges of each indicator to determine the social risk level of the target university student. The judgment method is as follows: If any indicator is in the high threshold range, the target college student is judged to have severe anxiety; if no indicator is in the high threshold range, and at least two indicators are in the medium threshold range, the student is judged to have moderate anxiety; if all indicators are in the low risk range or only one indicator is in the medium threshold range, the student is judged to have low anxiety.

6. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data according to claim 1, characterized in that, The method for generating risk profiles includes: A double-helix risk network is constructed, which includes an input layer, a spiral channel layer, a coupling engine layer, and a label generation layer. The spiral channel layer includes the anxiety spiral channel and the behavior spiral channel. The anxiety spiral channel forms a time series graph of the social anxiety prediction score sequence received by the input layer and determines the fluctuation feature vector; the behavior spiral channel forms a behavioral psychological tensor of the behavioral-psychological association evolution pattern sequence received by the input layer and determines the behavioral cluster center trajectory. The coupling engine layer obtains the coupling feature set from the fluctuation feature vector and the behavior cluster center trajectory. The coupling feature set includes the mean of the behavior cluster center, the time-frequency joint feature and the synchronization jump component. The label generation layer outputs a triple risk label based on the coupled feature set. The triple risk label includes the dominant behavioral pattern, anxiety score fluctuation type, and jump sensitivity category label. A risk profile is formed based on the triple risk label.

7. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data according to claim 1, characterized in that, The method for determining the set of candidate intervention strategies includes: A multi-level intervention response framework is constructed, which divides response levels according to social risk level. Each response level contains several intervention programs, and each intervention program is associated with a set of adaptation condition vectors, including dominant behavioral patterns, anxiety score fluctuation types, and change sensitivity categories. The corresponding response levels are associated with the social risk level of the target college students, and the matching degree score is obtained by matching the adaptation condition vectors of each intervention program in the corresponding response level with the risk profile of the target college students. Set a matching threshold, filter out intervention programs with matching scores greater than the threshold, and prioritize them according to matching scores to determine the candidate intervention program set for the target college students.

8. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data according to claim 7, characterized in that, The method for determining the optimal intervention plan includes: Obtain the reward and punishment signals of each intervention plan in the candidate intervention plan set during the historical push process; dynamically adjust the priority of each intervention plan in the candidate intervention plan set based on the reward and punishment signals, determine the intervention plan with the highest priority as the optimal intervention plan, and implement the intervention on the target college students based on the optimal intervention plan.

9. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data according to claim 1, characterized in that, The method for analyzing the feedback data includes: The study aims to obtain feedback data from target college students after implementing the optimal intervention plan. This data includes data on click behavior on intervention content, dwell time on intervention pages, and completion status of intervention guidance tasks. A quantitative score is used to determine the effectiveness of the feedback. An effective threshold is set, and feedback data with effectiveness scores above the threshold are designated as reward signals, while those below are designated as punishment signals.

10. The method for predicting and intervening in the depth of social anxiety tendencies among college students by integrating behavioral data according to claim 9, characterized in that, The method for dynamically adjusting the selection logic of the optimal intervention plan includes: If the reward / punishment signal corresponding to the intervention plan is a reward signal, the priority of the intervention plan in the candidate intervention strategy set will be increased by one level; conversely, if the reward / punishment signal is a punishment signal, the priority of the intervention plan in the candidate intervention strategy set will be decreased by one level.

Images