Student psychological assessment early warning method and system based on group behavior clustering analysis

Abstract

The invention discloses a student psychological assessment early warning method and system based on group behavior clustering analysis, and the method comprises the steps: obtaining multi-source campus behavior time series data, extracting group behavior feature vectors, and dynamically dividing a plurality of behavior normal form groups; calculating a multi-dimensional behavior baseline of each behavior normal form group; calculating a group relative deviation degree between the current behavior feature vector of the target student and the multi-dimensional behavior baseline of the behavior normal form group to which the target student belongs; analyzing the continuous increasing trend of the relative deviation degree of the group based on a sliding time window to judge adaptive behavior deviation; and performing attribution analysis on the target student with adaptive behavior deviation, generating a semantic risk early warning signal, and outputting targeted assistance suggestions according to the association intervention strategy library. According to the invention, through dynamic group division and relative deviation analysis, accurate and explainable early warning and effective intervention guidance of student behavior risks are realized.

Description

Technical Field

[0001] This invention relates to the field of mental health monitoring technology, specifically to a student psychological assessment and early warning method and system based on group behavior clustering analysis. Background Technology

[0002] Student mental health is a key focus of higher education management. Traditional assessment and early warning systems rely primarily on subjective or periodic methods such as counselor observation and psychological scale testing, which suffer from narrow coverage and delayed effectiveness. With the advancement of smart campus construction, automated analysis and early warning using various behavioral data generated by students on campus (such as consumption, entry / exit, and internet usage) has become an important technological development direction. Existing solutions typically identify individuals with abnormal behavior by constructing data analysis models that compare individual student behavioral data with preset "normal" patterns or group average levels.

[0003] However, such methods face a fundamental challenge in practical application: students' daily behavioral patterns are inherently diverse and group-oriented. Simply using a uniform "standard" to measure all students often results in insufficient accuracy and interpretability of the warning results. On the one hand, the model struggles to distinguish between normal behaviors belonging to a specific student group and abnormal behavioral changes that truly indicate risk, easily leading to misjudgments. On the other hand, even when abnormalities are identified, they often only provide a general risk warning, failing to clearly reveal the specific dimensions and evolutionary trends of behavioral deviations, resulting in a lack of precise entry points for subsequent intervention and support. Summary of the Invention

[0004] In view of the above problems, the present invention provides a student psychological assessment and early warning method and system based on group behavior clustering analysis. By dynamically dividing behavioral paradigm groups and calculating the continuous deviation trend of individuals relative to their respective groups, the present invention achieves interpretable and accurate early warning of students' behavioral adaptation risks.

[0005] To achieve the above objectives, in a first aspect, this application provides a student psychological assessment and early warning method based on group behavior clustering analysis, comprising:

[0006] Acquire multi-source time-series data on campus behavior of the target student group;

[0007] Based on multi-source campus behavior time series data, group behavior feature vectors are extracted, and unsupervised clustering algorithm is used to dynamically divide multiple behavioral paradigm groups.

[0008] Calculate the multidimensional behavioral baseline for each behavioral paradigm group. The multidimensional behavioral baseline represents the range of normal behavioral patterns of the behavioral paradigm group within a preset time period.

[0009] Calculate the relative deviation between the target student’s current behavioral feature vector and the multidimensional behavioral baseline of the behavioral paradigm group to which it historically belongs;

[0010] Based on a sliding time window, a persistent trend analysis is performed on the relative deviation of the group. When a statistically significant persistent increasing trend in the relative deviation of the group is detected, it is determined that an adaptive behavior deviation has occurred.

[0011] For target students who exhibit adaptive behavioral deviations, conduct deviation dimension attribution analysis to identify the specific behavioral dimensions that dominate the deviations, and generate risk warning signals with semantic types by combining the changing patterns of their behavioral paradigm group affiliation.

[0012] Based on risk warning signals, and in conjunction with a pre-set intervention strategy library, targeted assistance and guidance suggestions are generated and output.

[0013] Furthermore, based on multi-source campus behavior time-series data, group behavior feature vectors are extracted, and an unsupervised clustering algorithm is used to dynamically divide groups into multiple behavioral paradigms, including:

[0014] From multi-source campus behavior time-series data, spatiotemporal trajectory features, consumption pattern features and social collaboration features are extracted according to a preset period. Spatiotemporal trajectory features include the time distribution of regular entry and exit from core teaching and living places. Consumption pattern features include the proportion of consumption level and consumption structure. Social collaboration features are calculated based on the frequency of common occurrence at the same spatiotemporal point.

[0015] The spatiotemporal trajectory features, consumption pattern features, and social collaboration features are standardized to eliminate differences in units. They are then weighted based on the information entropy of the features or predefined importance weights. The weighted feature values ​​of each dimension are then concatenated to construct a unified multidimensional behavioral feature vector for each student.

[0016] Using a set of multidimensional behavioral feature vectors as input, unsupervised learning is performed using a density-based clustering algorithm. The density-based clustering algorithm identifies core points, boundary points, and noise points by setting a neighborhood radius and a minimum number of points threshold, and connects mutually reachable core points and boundary points in their neighborhoods to form multiple density-connected sample sets.

[0017] Each density-connected set of samples is defined as a behavioral paradigm group, and the central feature vector of each behavioral paradigm group is calculated.

[0018] The process of feature extraction, weighted fusion, and density-based clustering is periodically repeated to update the membership composition and central feature vector of the behavioral paradigm group, thereby achieving dynamic partitioning of the behavioral paradigm group.

[0019] Furthermore, the multidimensional behavioral baseline for each behavioral paradigm group is calculated, including:

[0020] For each behavioral paradigm group, obtain the multidimensional behavioral feature vector sequence of all its members within a preset historical time period;

[0021] For each feature dimension in the multidimensional behavioral feature vector sequence, calculate the statistical distribution of each feature dimension, including the mean and variance;

[0022] Based on the statistical distribution of each feature dimension, a multidimensional behavioral baseline interval is constructed. The multidimensional behavioral baseline interval is jointly defined by the normal value range of each feature dimension. The normal value range is determined by the confidence interval consisting of the mean plus or minus a certain number of variances.

[0023] The multidimensional behavioral baseline interval is associated with and stored with the central feature vector of the behavioral paradigm group, serving as a quantitative reference benchmark to characterize the normal behavioral pattern of the behavioral paradigm group.

[0024] Furthermore, the relative deviation between the target student's current behavioral feature vector and the multidimensional behavioral baseline of their historical behavioral paradigm group is calculated, including:

[0025] Identify the behavioral paradigm group to which the target student belonged with the highest intensity before the current analysis point, and use this as the historical behavioral paradigm group to which the student belonged.

[0026] Obtain the multidimensional behavioral baseline interval and its central feature vector of the historical behavioral paradigm group;

[0027] The original deviation vector is obtained by calculating the difference between the multidimensional behavioral feature vector of the target student at the current analysis time point and the central feature vector in each dimension.

[0028] For each dimension of the original deviation vector, the deviation value is normalized using the normal value range of the corresponding dimension in the multidimensional behavior baseline interval to obtain the standardized deviation value.

[0029] The standardized deviation values ​​are weighted and summed according to the preset dimension weights to calculate the Mahalanobis distance, which is used as the initial deviation.

[0030] The initial deviation is compared with the preset deviation threshold. If the initial deviation exceeds the deviation threshold, it is determined that the target student has exceeded the normal behavior pattern of the historical behavioral paradigm group, and the initial deviation is used as the output value of the group's relative deviation.

[0031] If the initial deviation does not exceed the deviation threshold, the relative deviation of the group will be output as zero.

[0032] Furthermore, for each dimension of the original deviation vector, the deviation value is normalized using the normal value range of the corresponding dimension in the multidimensional behavioral baseline interval to obtain the standardized deviation value, including:

[0033] For any dimension of the deviation value in the original deviation vector, obtain the upper limit and lower limit of the normal value range of the corresponding dimension in the multidimensional behavior baseline interval;

[0034] Calculate the difference between the upper limit and the lower limit of the normal value range, and use it as the length of the normal value range for this dimension;

[0035] Determine the sign of the dimension deviation value. If the dimension deviation value is positive, it indicates that the behavioral feature value of the target student in that dimension is higher than the corresponding value of the central feature vector.

[0036] If the dimension deviation value is positive, then divide the dimension deviation value by the difference between the upper limit of the normal value range and the corresponding value of the central feature vector to obtain the positive standardized deviation value of that dimension.

[0037] If the dimension deviation value is negative, it indicates that the behavioral feature value of the target student in that dimension is lower than the corresponding value of the central feature vector;

[0038] If the dimension deviation value is negative, the absolute value of the dimension deviation value is divided by the difference between the corresponding value of the central feature vector and the lower limit of the normal value range to obtain the negative standardized deviation value of that dimension.

[0039] The positive or negative standardized deviation value is output as the standardized deviation value for that dimension.

[0040] Furthermore, based on a sliding time window, a continuous trend analysis of the relative deviation of the group is conducted, including:

[0041] Using the current analysis time point as a baseline, a sliding time window is formed by extracting a sequence of relative deviations of the population that includes multiple consecutive time points.

[0042] For the population relative deviation sequence within the sliding time window, linear regression is used to fit the trend line of the population relative deviation sequence, and the slope of the trend line is calculated.

[0043] For the population relative deviation sequence within the sliding time window, the cumulative sum control algorithm is applied to calculate the cumulative sum statistic at each time point. The cumulative sum statistic reflects the cumulative deviation of the sequence value relative to the historical mean.

[0044] Set the trend slope threshold and cumulative and statistical control limits for the trend line;

[0045] If the slope of the trend line is greater than the trend slope threshold, and the cumulative sum statistic continues to exceed the cumulative sum statistic control limit, then the relative deviation of the group is judged to show a statistically significant and continuous increasing trend.

[0046] Record the starting point and current intensity of the persistent increasing trend as temporal evidence for determining the occurrence of adaptive behavioral deviations.

[0047] Furthermore, for target students exhibiting adaptive behavioral deviations, attribution analysis of deviation dimensions is conducted to identify the specific behavioral dimensions that dominate the deviation. Combined with changes in their behavioral paradigm group affiliation patterns, semantically typed risk warning signals are generated, including:

[0048] Extract the feature dimensions whose contribution exceeds a preset contribution threshold when calculating the relative deviation of the group, and use them as the dominant deviation dimension;

[0049] Based on the feature category to which the dominant deviation dimension belongs, the main aspects of behavioral deviation are determined. Feature categories include spatiotemporal trajectory, consumption patterns, and social collaboration.

[0050] Calculate the similarity between the target student and the central feature vector of each behavioral paradigm group at the current analysis time point and at historical time points, and denote it as the target student similarity.

[0051] Identify the behavioral paradigm group with the highest historical affiliation strength of the target students as the source group, and identify the behavioral paradigm group with the fastest current similarity growth to the target students or that has exceeded the preset similarity threshold as the potential approaching group;

[0052] The main aspects of behavioral deviation, the type attributes of the source group, and the type attributes of the potential approaching group are combined and mapped.

[0053] Based on a predefined mapping rule library, corresponding semantic type labels are assigned to combinations. Semantic type labels include social isolation, interest shift, behavioral disorder, or economic pressure.

[0054] The information of semantic type labels, dominant deviation dimensions and their contribution, source group and potential similar groups is encapsulated to generate risk warning signals.

[0055] Furthermore, the main aspects of behavioral deviation, the type attributes of the source group, and the type attributes of potential approaching groups are combined and mapped; based on a predefined mapping rule library, corresponding semantic type labels are assigned to the combination, including:

[0056] The main aspects of behavioral deviation, the type attributes of the source group, and the type attributes of the potential approaching group are encoded into triplet data.

[0057] The triplet data is matched with the rule conditions in the predefined mapping rule base. Each rule in the mapping rule base contains a condition part and a conclusion part. The condition part defines a specific combination pattern of behavioral deviation from the main aspect, source group type attribute and potential approaching group type attribute.

[0058] When the triplet data satisfies part of the condition of a rule, that rule is triggered;

[0059] The conclusion part of the triggered rule is executed. The operations in the conclusion part include directly outputting the predefined semantic type label, or calling the scoring function to calculate the confidence of multiple candidate semantic type labels and selecting the label with the highest confidence to output.

[0060] The output semantic type label is associated with the triple data that triggered the label and the specific rule identifier that was matched, thus completing the combination mapping and label assignment.

[0061] Furthermore, based on risk warning signals, and in conjunction with a pre-set intervention strategy library, targeted support and guidance suggestions are generated and output, including:

[0062] Analyze risk warning signals and extract semantic type labels, dominant deviation dimension information, and source group and potential similar group information;

[0063] Using semantic type tags as the primary key, a search is performed in a pre-built intervention strategy library, which stores basic intervention strategy templates and historical success case indexes corresponding to different semantic type tags.

[0064] Based on the dominant deviation dimension information, specific intervention action descriptions with the highest correlation to the dominant deviation dimension are selected from the retrieved basic intervention strategy templates.

[0065] Based on information about the source group and potential similar groups, the description of specific intervention actions is parametrically adjusted. The parametric adjustments include suggested intervention personnel, suggested activity scenarios, and suggested communication entry points.

[0066] The parameterized descriptions of specific intervention actions are integrated with similar case summaries matched from the historical success case index to generate a structured draft of assistance and guidance suggestions;

[0067] The draft of the support and guidance recommendations is linked with the basic profile information of the target students, and a summary of key evidence for risk warning signals is attached. The final targeted support and guidance recommendation document is then generated and output.

[0068] In a second aspect, the present invention also provides a student psychological assessment and early warning system based on group behavior clustering analysis, applicable to the method described in the first aspect. The system includes a data acquisition and fusion module, a dynamic group clustering module, a group baseline calculation module, an individual deviation calculation module, a persistent trend analysis module, a risk signal generation module, and a support suggestion generation module. The data acquisition and fusion module is configured to acquire and integrate multi-source campus behavior time-series data of the target student group; the dynamic group clustering module is configured to extract group behavior feature vectors based on multi-source campus behavior time-series data and dynamically divide multiple behavioral paradigm groups using an unsupervised clustering algorithm; the group baseline calculation module is configured to calculate a multi-dimensional behavioral baseline for each behavioral paradigm group, the multi-dimensional behavioral baseline representing the normal behavioral pattern range of the behavioral paradigm group within a preset time period. The individual deviation calculation module is configured to calculate the relative deviation between the target student's current behavioral feature vector and the multidimensional behavioral baseline of their historical behavioral paradigm group. The persistent trend analysis module is configured to perform persistent trend analysis on the relative deviation of the group based on a sliding time window. When a statistically significant persistent increasing trend in the relative deviation of the group is detected, adaptive behavioral deviation is determined to have occurred. The risk signal generation module is configured to perform deviation dimension attribution analysis on the target students who have experienced adaptive behavioral deviation, identify the specific behavioral dimensions that dominate the deviation, and generate a semantically typed risk warning signal in combination with the changing patterns of their behavioral paradigm group affiliation. The support suggestion generation module is configured to generate and output targeted support and guidance suggestions based on the risk warning signal, in conjunction with a pre-set intervention strategy library.

[0069] Unlike existing technologies, the above-mentioned technical solution provides a student psychological assessment and early warning method and system based on group behavior clustering analysis. This includes: acquiring multi-source campus behavior time-series data, extracting group behavior feature vectors, and dynamically dividing multiple behavioral paradigm groups; calculating the multidimensional behavioral baseline of each behavioral paradigm group; calculating the relative deviation between the target student's current behavioral feature vector and the multidimensional behavioral baseline of its historical behavioral paradigm group; analyzing the continuous increasing trend of this relative deviation based on a sliding time window to determine adaptive behavioral deviation; performing attribution analysis on target students exhibiting adaptive behavioral deviation, generating semantic risk warning signals, and outputting targeted support suggestions based on an intervention strategy library. This invention, through dynamic group segmentation and relative deviation analysis, achieves accurate, interpretable early warning and effective intervention guidance for student behavioral risks.

[0070] The above description of the invention is merely an overview of the technical solution of the present invention. In order to enable those skilled in the art to better understand the technical solution of the present invention and to implement it based on the description and drawings, and to make the above-mentioned objectives and other objectives, features and advantages of the present invention easier to understand, the following description is provided in conjunction with the specific embodiments and drawings of the present invention. Attached Figure Description

[0071] The accompanying drawings are only used to illustrate the principles, implementation methods, applications, features, and effects of specific embodiments of the present invention and other related contents, and should not be considered as limitations on the present invention.

[0072] In the accompanying drawings of the instruction manual:

[0073] Figure 1 This is a schematic diagram illustrating steps S101 to S107 of the method described in the specific implementation embodiment;

[0074] Figure 2 This is a schematic diagram illustrating steps S201 to S205 of the method described in a specific implementation.

[0075] Figure 3 This is a schematic diagram illustrating steps S301 to S306 of the method described in a specific implementation.

[0076] Figure 4 This is a schematic diagram illustrating steps S401 to S407 of the method described in a specific embodiment;

[0077] Figure 5 This is a schematic diagram of the student psychological assessment and early warning system described in a specific implementation.

[0078] The reference numerals used in the above figures are explained as follows:

[0079] 1. Student psychological assessment and early warning system;

[0080] 11. Data Acquisition and Fusion Module;

[0081] 12. Dynamic group clustering module;

[0082] 13. Population baseline calculation module;

[0083] 14. Individual Deviation Calculation Module;

[0084] 15. Continuous Trend Analysis Module;

[0085] 16. Risk signal generation module;

[0086] 17. Assistance suggestion generation module. Detailed Implementation

[0087] To illustrate the possible application scenarios, technical principles, implementable specific solutions, and achievable objectives and effects of this invention in detail, the following description, in conjunction with the listed specific embodiments and accompanying drawings, provides a detailed explanation. The embodiments described herein are merely illustrative of the technical solutions of this invention and are therefore intended only as examples, not as limiting the scope of protection of this invention.

[0088] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this invention, as long as there are no technical contradictions or conflicts, the technical features mentioned in each embodiment can be combined in any way to form corresponding implementable technical solutions.

[0089] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit the invention.

[0090] In the description of this invention, the term "and / or" is used to describe the logical relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A exists, B exists, and A and B exist simultaneously. Additionally, the character " / " generally indicates that the preceding and following objects have an "or" logical relationship.

[0091] In this invention, terms such as “first” and “second” are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any actual quantity, hierarchy, or order between these entities or operations.

[0092] Without further limitations, the use of terms such as “comprising,” “including,” “having,” or other similar open-ended expressions in this invention is intended to cover non-exclusive inclusion, which does not exclude the presence of additional elements in a process, method, or product that includes the stated elements, such that a process, method, or product that includes a list of elements may include not only those defined elements but also other elements not expressly listed, or elements inherent to such a process, method, or product.

[0093] Similar to the understanding in the Examination Guidelines, in this invention, expressions such as "greater than," "less than," and "exceeding" are understood to exclude the stated number; expressions such as "above," "below," and "within" are understood to include the stated number. Furthermore, in the description of the embodiments of this invention, "multiple" means two or more (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups" and "multiple times," unless otherwise explicitly specified.

[0094] In the description of the embodiments of the present invention, the spatial related expressions used, such as "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "vertical," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," "circumferential," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the specific embodiments or drawings. They are only for the purpose of describing the specific embodiments of the present invention or for the reader's understanding, and do not indicate or imply that the device or component referred to must have a specific position, a specific orientation, or be constructed or operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention.

[0095] Unless otherwise explicitly stated or limited, the terms "installation," "connection," "linking," "fixing," and "setting," as used in the description of the embodiments of this invention, should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral arrangement; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this invention according to the specific circumstances.

[0096] Please see Figure 1 In a first aspect, this embodiment provides a student psychological assessment and early warning method based on group behavior clustering analysis, including:

[0097] S101. Obtain multi-source campus behavior time-series data of the target student group;

[0098] S102. Based on multi-source campus behavior time series data, extract group behavior feature vectors and use unsupervised clustering algorithm to dynamically divide multiple behavior paradigm groups;

[0099] S103. Calculate the multidimensional behavioral baseline for each behavioral paradigm group. The multidimensional behavioral baseline represents the range of normal behavioral patterns of the behavioral paradigm group within a preset time period.

[0100] S104. Calculate the relative deviation between the target student’s current behavioral feature vector and the multidimensional behavioral baseline of the behavioral paradigm group to which it historically belongs;

[0101] S105. Based on the sliding time window, perform a continuous trend analysis on the relative deviation of the group. When a statistically significant continuous increasing trend in the relative deviation of the group is detected, it is determined that an adaptive behavior deviation has occurred.

[0102] S106. For target students who exhibit adaptive behavioral deviations, conduct deviation dimension attribution analysis to identify the specific behavioral dimensions that dominate the deviations, and generate risk warning signals with semantic types by combining the changing patterns of their behavioral paradigm group affiliation.

[0103] S107. Based on the risk warning signal, associate it with the pre-set intervention strategy library to generate and output targeted assistance and guidance suggestions.

[0104] In step S101, multi-source campus behavior time-series data is automatically collected and integrated through interfaces with various information management systems within the campus. These data sources include at least the campus card system that records consumption and access control information, the authentication and billing system that reflects network access behavior, and the academic affairs management system that records courses and grades. The collected raw data is aligned using student ID numbers as unique identifiers and aggregated according to a unified time granularity to form a structured, timestamped sequence of individual student behaviors, providing a data foundation for subsequent time-series pattern analysis.

[0105] In step S102, multi-dimensional features that can quantify student behavior patterns are extracted from multi-source campus behavior time-series data at preset periods. These features include the distribution of entry and exit times to characterize daily routines, statistics on the amount and category of purchases to reflect consumption habits, and social collaboration calculated based on the frequency of shared occurrences in the same time and space. These features are standardized and weighted to create a group behavior feature vector for each student. An unsupervised clustering algorithm is used to analyze this feature vector set, automatically dividing students into multiple sets with similar internal patterns based on the similarity between vectors. Each set is defined as a behavioral paradigm group. Furthermore, to achieve dynamic partitioning, the system periodically re-executes the feature extraction and clustering process, allowing the composition of behavioral paradigm groups to update along with the evolution of the overall student behavior patterns.

[0106] In step S103, for each segmented behavioral paradigm group, the feature vector sequences of all its members within a preset historical time period are collected. Statistical analysis is performed on each feature dimension in the sequence to obtain the central tendency and dispersion indices for each dimension. Based on these statistical indices, a numerical interval characterizing the normal fluctuation range of the group is defined for each feature dimension. The normal intervals of all dimensions jointly constitute the multidimensional behavioral baseline of the group. This multidimensional behavioral baseline serves as a quantitative reference to describe the stable and typical behavioral patterns of the group within the historical time period.

[0107] In step S104, the behavioral paradigm group to which the target student primarily belonged before the current analysis point is determined as its historical behavioral paradigm group. A multidimensional behavioral baseline of this behavioral paradigm group is obtained and compared with the current behavioral feature vector generated by the target student based on recent data. The relative deviation of the group is obtained by calculating the comprehensive degree of deviation of the current feature vector from the historical group baseline; this indicator quantifies the degree to which an individual's behavioral pattern deviates from the norm of their community.

[0108] In step S105, since a single increase in the relative deviation of the group may be caused by accidental factors, the relative deviation of the target students calculated at different analysis time points is continuously recorded to form a continuous record of the indicator arranged chronologically. Based on a sliding analysis window containing multiple consecutive time points, statistical trend analysis is performed on these continuous records within the window to identify whether the sequence has a statistically significant, monotonically increasing change pattern. When such a continuous increasing trend is detected, it is determined that the target student has undergone adaptive behavioral deviation, indicating that their behavior pattern is continuously deviating from the normality of their original community.

[0109] In step S106, for students exhibiting adaptive behavioral deviations, the composition of their relative deviation within their group is analyzed, and the feature dimension contributing the most is identified as the specific behavioral dimension of the dominant deviation. Simultaneously, changes in the similarity between the student and various behavioral paradigm groups are tracked, and the changing patterns of their group affiliation are analyzed to identify potential source groups they may leave and potential groups they may converge towards. Combining the dominant deviation dimension and changes in affiliation, a predefined mapping mechanism is preferably used to generate a risk warning signal with semantic type, thus generating a risk warning signal containing specific attribution information.

[0110] In step S107, a pre-built intervention strategy library is constructed based on historical intervention experience and expert knowledge. It stores suggested intervention measures, communication strategies, or activity plans corresponding to risk signals of different semantic types. The generated risk warning signals are parsed, and their semantic types and other key information are used as indexes to query the pre-built intervention strategy library. Based on the matched strategy templates and the specific behavioral performance of the target student (such as the specific value of the dominant deviation dimension), parameterized filling and adaptation are performed, ultimately generating and outputting structured, actionable support and guidance suggestions.

[0111] This embodiment constructs a student psychological assessment and early warning process from multi-source data collection to the generation of targeted intervention suggestions. It dynamically identifies student behavioral paradigm groups using an unsupervised clustering algorithm and establishes a multi-dimensional behavioral baseline for each paradigm group. The persistent relative deviation of an individual from the multi-dimensional behavioral baseline of their respective paradigm group is then used as a risk criterion. Semantic risk warning signals are generated through deviation dimension attribution analysis. This embodiment transforms the assessment standard from absolute thresholds to relative baselines and achieves interpretability and operability of the early warning results, effectively improving the accuracy and practicality of campus psychological risk early warning.

[0112] Please see Figure 2 In some embodiments, based on multi-source campus behavior time-series data, group behavior feature vectors are extracted, and unsupervised clustering algorithms are used to dynamically divide multiple behavioral paradigm groups, including:

[0113] S201. Extract spatiotemporal trajectory features, consumption pattern features, and social collaboration features from multi-source campus behavior time series data according to a preset period. The spatiotemporal trajectory features include the time distribution of regular entry and exit from core teaching and living places. The consumption pattern features include the consumption level and consumption structure ratio. The social collaboration features are calculated based on the frequency of common occurrence at the same spatiotemporal point.

[0114] S202. The spatiotemporal trajectory features, consumption pattern features and social collaboration features are standardized to eliminate the difference in units. They are then weighted based on the information entropy of the features or predefined importance weights. The weighted feature values ​​of each dimension are then concatenated to construct a unified multidimensional behavioral feature vector for each student.

[0115] S203. Using a set of multidimensional behavioral feature vectors as input, unsupervised learning is performed using a density-based clustering algorithm. The density-based clustering algorithm identifies core points, boundary points, and noise points by setting a neighborhood radius and a minimum number of points threshold, and connects mutually reachable core points and boundary points in their neighborhoods to form multiple density-connected sample sets.

[0116] S204. Define each density-connected sample set as a behavioral paradigm group, and calculate the central feature vector of each behavioral paradigm group.

[0117] S205. Periodically repeat the feature extraction, weighted fusion and density-based clustering process to update the membership composition and central feature vector of the behavioral paradigm group, thereby realizing the dynamic division of the behavioral paradigm group.

[0118] In step S201, the spatiotemporal trajectory features are extracted by analyzing the regular entry and exit times of students in core teaching locations (such as teaching buildings and libraries) and living locations (such as dormitories and canteens) recorded in the student ID card and access control system. For example, the frequency distribution of students entering and exiting these locations at different times can be statistically analyzed. The consumption pattern features are derived from the student ID card transaction records. The consumption level is obtained by calculating the total consumption amount and average daily consumption amount within a preset period, and the consumption structure ratio is obtained by statistically analyzing the proportion of expenditure in different consumption categories (such as catering, shopping, and entertainment). The social collaboration features are calculated by analyzing the frequency of different students appearing in the same location at similar time points (such as setting a time tolerance window) (e.g., consuming at the same canteen window at the same time, or entering and exiting the same library at the same time). This frequency value can characterize the synchronicity of behavior or the strength of social association among students.

[0119] In step S202, the standardization process can employ the Z-score standardization method to transform each feature value into a distribution with a mean of 0 and a standard deviation of 1, thereby eliminating the influence of differences in the units and numerical ranges of different features. The weighting process can determine weights based on the information entropy of the features; features with low information entropy (low uncertainty, high discriminative power) are assigned higher weights. Alternatively, importance weights can be predefined based on prior business knowledge; for example, if social collaboration is considered more crucial for psychological assessment, it is assigned a higher weight. The weighted feature values ​​of each dimension are then concatenated in a fixed order to construct a unified multidimensional behavioral feature vector for each student. This vector serves as the input unit for subsequent clustering analysis.

[0120] In step S203, the neighborhood radius can be determined by statistical analysis of the multidimensional behavioral feature vector set, for example, by calculating the average or median distance between all sample points and multiplying it by an empirical coefficient. For any sample point in space (i.e., a student's feature vector), the number of sample points contained within the spatial range centered on that point and with the neighborhood radius as its radius is called the neighborhood density of that point. If the neighborhood density of a sample point is not less than the minimum number of points threshold, then the point is marked as a core point; if the neighborhood density of a sample point is less than the minimum number of points threshold, but it is located within the neighborhood of a core point, then the point is marked as a boundary point; points that are neither core points nor within the neighborhood of any core point are marked as noise points. The algorithm forms clusters by finding density-connected sample sets: starting from a core point, all core points that are density-reachable (connected through a series of core point neighborhoods) and their neighborhood boundary points together constitute a density-connected sample set.

[0121] In step S204, the central feature vector of each behavioral paradigm group is calculated. Typically, the mean vector of the multidimensional behavioral feature vectors of all members in the group can be taken as its center to characterize the typical behavioral pattern of the group.

[0122] In step S205, the process is periodically triggered at fixed time intervals (e.g., monthly or per semester) to restart from the latest multi-source campus behavior time-series data and execute the feature extraction, weighted fusion, and density-based clustering processes described in steps S201 to S204, thereby dynamically dividing the behavioral paradigm groups. Through this periodic recalculation, the membership composition and central feature vectors of the behavioral paradigm groups are updated, reflecting the natural evolution of student behavior patterns over time, the formation of new groups, or the dissipation of old groups, ensuring that the group reference system upon which the entire early warning system relies is currently valid.

[0123] This embodiment defines multi-dimensional behavioral characteristics and uses a density-based clustering algorithm to objectively discover naturally formed behavioral paradigm groups with inherent similarities on campus, and can effectively handle noisy data (i.e., students who do not belong to any stable behavioral paradigm group). The regular update mechanism ensures the timeliness of group segmentation and provides an accurate and dynamic reference benchmark for subsequent risk assessment based on relative deviation of groups.

[0124] In some embodiments, calculating a multidimensional behavioral baseline for each behavioral paradigm group includes:

[0125] For each behavioral paradigm group, obtain the multidimensional behavioral feature vector sequence of all its members within a preset historical time period;

[0126] For each feature dimension in the multidimensional behavioral feature vector sequence, calculate the statistical distribution of each feature dimension, including the mean and variance;

[0127] Based on the statistical distribution of each feature dimension, a multidimensional behavioral baseline interval is constructed. The multidimensional behavioral baseline interval is jointly defined by the normal value range of each feature dimension. The normal value range is determined by the confidence interval consisting of the mean plus or minus a certain number of variances.

[0128] The multidimensional behavioral baseline interval is associated with and stored with the central feature vector of the behavioral paradigm group, serving as a quantitative reference benchmark to characterize the normal behavioral pattern of the behavioral paradigm group.

[0129] In this embodiment, the preset historical time period is a configurable parameter, such as the most recent two months or the previous semester, used to obtain a sufficient amount of historical data that reflects the recent stable behavioral patterns of the group. For each behavioral paradigm group, the multidimensional behavioral feature vectors of each member in the preset historical time period are obtained from its member list, arranged in chronological order. The set of these vectors constitutes the multidimensional behavioral feature vector sequence of the group.

[0130] For the obtained multidimensional behavioral feature vector sequence, calculations are performed for all values ​​in each feature dimension. The mean reflects the average level of group behavior in that dimension, while the variance characterizes the dispersion of group members' behavior in that dimension. The mean and variance together constitute the statistical distribution describing the distribution of that feature dimension within the group.

[0131] Based on the calculated mean and variance of each feature dimension, a normal range for that dimension is constructed. Specifically, the normal range can be determined by adding or subtracting a certain multiple of the variance from the mean to form a confidence interval; this multiple is a positive number set according to the required confidence level. The normal ranges of all feature dimensions collectively define a region in a multidimensional space, which is the multidimensional behavioral baseline interval for this behavioral paradigm group. This interval quantitatively characterizes the normal range in which the behavioral feature values ​​of most members of this group fall within a preset historical time period.

[0132] The calculated multidimensional behavioral baseline interval is associated and stored with the central feature vector of the behavioral paradigm group. Specifically, this can be achieved by storing the upper and lower bound vectors of the baseline interval, the group identifier, and the central feature vector in the same record in the database. The stored multidimensional behavioral baseline interval serves as a quantitative reference benchmark for subsequent steps to determine whether an individual's behavior deviates from the norm of their group.

[0133] This embodiment uses statistical methods to expand the behavioral pattern of the behavioral paradigm group from the center point to a reasonable fluctuation range. This ranged baseline can better accommodate normal differences within the group, providing a robust reference standard for subsequent calculation of the deviation of an individual from the group baseline, so as to ensure the accuracy of the early warning method and reduce false alarms.

[0134] Please see Figure 3 In some embodiments, the relative deviation between the target student's current behavioral feature vector and the multidimensional behavioral baseline of its historical behavioral paradigm group is calculated, including:

[0135] S301. Identify the behavioral paradigm group with the highest affiliation intensity for the target student before the current analysis point in time, and use it as the historical behavioral paradigm group to which the student belongs.

[0136] S302. Obtain the multidimensional behavioral baseline interval and its central feature vector of the historical behavioral paradigm group;

[0137] S303. Calculate the difference between the multidimensional behavioral feature vector of the target student at the current analysis time point and the central feature vector in each dimension to obtain the original deviation vector.

[0138] S304. For each dimension deviation value in the original deviation vector, normalize it using the normal value range of the corresponding dimension in the multidimensional behavior baseline interval to obtain the standardized deviation value.

[0139] S305. The standardized deviation values ​​are weighted and summed according to the preset dimension weights to calculate the Mahalanobis distance, which is used as the initial deviation.

[0140] S306. Compare the initial deviation with the preset deviation threshold. If the initial deviation exceeds the deviation threshold, it is determined that the target student has exceeded the normal behavior pattern of the historical behavioral paradigm group, and the initial deviation is used as the output value of the group's relative deviation.

[0141] If the initial deviation does not exceed the deviation threshold, the relative deviation of the group will be output as zero.

[0142] In step S301, the affiliation strength is used to quantify the degree of affiliation between the target student and each behavioral paradigm group. Specifically, it can be obtained by calculating the similarity between the target student's historical behavioral feature vector and the central feature vector of each group (such as cosine similarity or the reciprocal of Euclidean distance), and the group with the highest similarity is taken as its historical behavioral paradigm group. In some other optional embodiments, it can also be determined based on the student's direct affiliation record in the historical clustering results.

[0143] In step S302, the multidimensional behavioral baseline interval and its central feature vector corresponding to the historical behavioral paradigm group determined in step S301 are retrieved and read from the stored behavioral paradigm group information. The multidimensional behavioral baseline interval includes the upper and lower limits of the normal value range of each feature dimension of the behavioral paradigm group, while the central feature vector represents the typical behavioral pattern of the behavioral paradigm group.

[0144] In step S303, the multidimensional behavioral feature vector of the target student at the current analysis time point is extracted from the behavioral data of its most recent analysis period (e.g., the current week). The multidimensional behavioral feature vector at the current analysis time point is subtracted element-wise from the central feature vector obtained in step S302, that is, the values ​​in the corresponding dimensions are subtracted. The resulting vector is the original deviation vector, where each value reflects the absolute deviation of the target student from the typical value of the group in the corresponding behavioral dimension.

[0145] In step S304, the deviation values ​​of each dimension in the original deviation vector are normalized to eliminate the incomparability caused by differences in the units and normal fluctuation ranges of different feature dimensions. Specifically, for the deviation value of any dimension, the normal value range (i.e., upper and lower limits) of that dimension in the corresponding multidimensional behavioral baseline interval is obtained; the original deviation value is scaled using this normal value range so that the processed standardized deviation value can reflect the relative size of the original deviation with respect to the normal fluctuation range of that dimension. Thus, the deviations of different dimensions are transformed to a comparable scale.

[0146] In step S305, the preset dimension weights can be pre-set based on business importance or feature stability; for example, social collaboration features may be assigned higher weights. The standardized deviation values ​​of each dimension obtained in step S304 are multiplied by their corresponding preset dimension weights, and then the sum of squares of all weighted values ​​is taken, followed by the square root, to calculate the Mahalanobis distance, which serves as the initial deviation. The Mahalanobis distance considers the correlation between features (if the covariance matrix between features is known), but in this simplified implementation, it is usually calculated based on the assumption of independence for each dimension. This distance value comprehensively reflects the overall deviation of the target student's current behavior from the center of the behavioral paradigm group in a multidimensional space.

[0147] In step S306, the preset deviation threshold is an empirical value or a critical value determined by statistical methods. For example, it can be based on the preliminary deviation distribution of a historical normal student group, taking a certain percentile as the threshold. The calculated preliminary deviation is compared with this threshold: if it exceeds the threshold, it indicates that the target student's behavior pattern has significantly deviated from the normal range of its historical group, and the value of the preliminary deviation is directly used as the final output of the group relative deviation; if it does not exceed the threshold, it is considered that its behavior is still within the normal range of the group, and the group relative deviation is output as zero, indicating that no significant deviation has occurred.

[0148] This embodiment determines the historical behavioral paradigm group based on the strength of attribution. Then, it normalizes the original deviation vector using the multidimensional behavioral baseline interval of this paradigm group. A preliminary deviation is obtained by calculating Mahalanobis distance, and the relative deviation of the group is output after comparison with a deviation threshold. This embodiment transforms the difference between individual behavior and group norms into a comparable and significant quantitative indicator, providing a precise data foundation for subsequent continuous trend analysis.

[0149] In some embodiments, for each dimension deviation value in the original deviation vector, normalization is performed using the normal value range of the corresponding dimension in the multidimensional behavioral baseline interval to obtain a standardized deviation value, including:

[0150] For any dimension of the deviation value in the original deviation vector, obtain the upper limit and lower limit of the normal value range of the corresponding dimension in the multidimensional behavior baseline interval;

[0151] Calculate the difference between the upper limit and the lower limit of the normal value range, and use it as the length of the normal value range for this dimension;

[0152] Determine the sign of the dimension deviation value. If the dimension deviation value is positive, it indicates that the behavioral feature value of the target student in that dimension is higher than the corresponding value of the central feature vector.

[0153] If the dimension deviation value is positive, then divide the dimension deviation value by the difference between the upper limit of the normal value range and the corresponding value of the central feature vector to obtain the positive standardized deviation value of that dimension.

[0154] If the dimension deviation value is negative, it indicates that the behavioral feature value of the target student in that dimension is lower than the corresponding value of the central feature vector;

[0155] If the dimension deviation value is negative, the absolute value of the dimension deviation value is divided by the difference between the corresponding value of the central feature vector and the lower limit of the normal value range to obtain the negative standardized deviation value of that dimension.

[0156] The positive or negative standardized deviation value is output as the standardized deviation value for that dimension.

[0157] In this embodiment, the sign of the dimensional deviation value indicates the direction of deviation of the target student behavioral feature value relative to the corresponding value of the central feature vector. The upper and lower limits of the normal value range are directly read from the stored multidimensional behavioral baseline interval.

[0158] The difference between the upper limit and the lower limit of the normal range is calculated, and the result is the length of the normal range for that dimension. This length quantifies the normal fluctuation range of the behavioral paradigm group in this dimension.

[0159] For any dimension deviation value in the original deviation vector, its sign is first determined. If the value is positive, it indicates that the target student's eigenvalue in this dimension is higher than the corresponding value in the central eigenvector. In this case, the dimension deviation value is divided by the difference between the upper limit of the normal range and the corresponding value in the central eigenvector, and the result is taken as the positive standardized deviation value for that dimension. This calculation allows the positive standardized deviation value to characterize the proportion of the current positive deviation in the allowed positive deviation space.

[0160] If the dimension bias value is negative, it indicates that the target student's feature value in this dimension is lower than the corresponding value of the central feature vector. In this case, the absolute value of the dimension bias value is divided by the difference between the corresponding value of the central feature vector and the lower limit of the normal range, and the result is used as the negative standardized bias value for that dimension. This calculation allows the negative standardized bias value to characterize the proportion of the current negative bias in the allowable negative deviation space.

[0161] Based on the sign of the dimensional deviation value, the corresponding positive or negative standardized deviation value is selected as the output standardized deviation value for that dimension. The output standardized deviation value is a dimensionless scalar. If its value is greater than 1, it indicates that the behavioral characteristics of the target student in that dimension have exceeded the normal range of the corresponding behavioral paradigm group.

[0162] This embodiment achieves dimensionless measurement and scale unification of deviation values ​​by comparing the original deviation values ​​with the allowable deviation space of the behavioral paradigm group in that dimension. By distinguishing between positive and negative directions and using a normalization method based on the allowable deviation space, the relative severity of the deviation relative to the normal range of the group can be accurately quantified, providing accurate and comparable basic data for subsequent comprehensive calculation of the relative deviation of the group.

[0163] In some embodiments, a continuous trend analysis of the relative deviation of the group is performed based on a sliding time window, including:

[0164] Using the current analysis time point as a baseline, a sliding time window is formed by extracting a sequence of relative deviations of the population that includes multiple consecutive time points.

[0165] For the population relative deviation sequence within the sliding time window, linear regression is used to fit the trend line of the population relative deviation sequence, and the slope of the trend line is calculated.

[0166] For the population relative deviation sequence within the sliding time window, the cumulative sum control algorithm is applied to calculate the cumulative sum statistic at each time point. The cumulative sum statistic reflects the cumulative deviation of the sequence value relative to the historical mean.

[0167] Set the trend slope threshold and cumulative and statistical control limits for the trend line;

[0168] If the slope of the trend line is greater than the trend slope threshold, and the cumulative sum statistic continues to exceed the cumulative sum statistic control limit, then the relative deviation of the group is judged to show a statistically significant and continuous increasing trend.

[0169] Record the starting point and current intensity of the persistent increasing trend as temporal evidence for determining the occurrence of adaptive behavioral deviations.

[0170] In this embodiment, the sliding time window is constructed by extracting a preset time length of relative deviation records of the group from the current analysis time point. This preset time length can cover a sufficient number of continuous analysis periods to ensure the statistical power of the trend analysis.

[0171] For the population relative deviation sequence constituting the sliding time window, the least squares method is used for linear regression fitting. With the time sequence number as the independent variable and the population relative deviation as the dependent variable, a straight line characterizing the trend of the sequence over time is calculated. The slope of this line is the slope of the trend line, and its value directly represents the average rate of change of the population relative deviation over time within the window period.

[0172] For sequences within the same sliding time window, a cumulative sum control algorithm is applied in parallel. A reference mean is calculated based on the initial portion of the sequence or historical stable period data. Then, for each data point within the window, the difference between it and the reference mean is calculated, and these differences are recursively accumulated and summed to obtain the cumulative sum statistic for each time point. This statistic has a cumulative amplifying effect on small but persistent upward shifts in sequence values ​​relative to the reference mean.

[0173] Trend slope threshold and cumulative sum control limits are pre-defined criteria based on statistical principles or historical data analysis. The trend slope threshold defines the boundary of the trend slope within the acceptable range of natural fluctuations; the cumulative sum control limits define the critical line for whether the cumulative offset is statistically significant.

[0174] To determine whether a group's relative deviation exhibits a statistically significant and persistently increasing trend, two conditions must be met simultaneously: the slope of the trend line obtained from linear regression exceeds a preset trend slope threshold, and the cumulative sum statistic calculated by the cumulative sum control algorithm remains consistently above a preset cumulative sum statistic control limit. This dual-condition determination mechanism combines trend direction and statistical significance testing. When these two conditions are met, the system records the starting point at which the trend was identified and the current trend strength index; this information together constitutes temporal evidence for determining the occurrence of adaptive behavioral deviation.

[0175] This embodiment quantifies the trend direction through linear regression and combines it with a cumulative sum control algorithm to detect statistically significant shifts, thereby achieving robust detection of the persistent deterioration trend of the relative deviation of the population. It not only identifies the direction of change but also confirms the persistence of the change through statistical testing, effectively filtering out random fluctuations and providing a reliable and timely time series judgment basis for subsequent risk warnings.

[0176] Please see Figure 4 In some embodiments, for target students exhibiting adaptive behavioral deviations, attribution analysis of deviation dimensions is performed to identify the specific behavioral dimensions that dominate the deviation. Combined with changes in their behavioral paradigm group affiliation patterns, a risk warning signal with semantic type is generated, including:

[0177] S401. Extract the feature dimension whose contribution exceeds the preset contribution threshold when calculating the relative deviation of the group, and use it as the dominant deviation dimension.

[0178] S402. Based on the feature category to which the dominant deviation dimension belongs, determine the main aspects of the behavioral deviation. The feature categories include spatiotemporal trajectory, consumption patterns, and social collaboration.

[0179] S403. Calculate the similarity between the target student and the central feature vector of each behavioral paradigm group at the current analysis time point and the historical time point, and denot it as the target student similarity.

[0180] S404. Identify the behavioral paradigm group with the highest historical affiliation strength of the target student as the source group, and identify the behavioral paradigm group with the fastest current similarity growth to the target student or that has exceeded the preset similarity threshold as the potential approaching group.

[0181] S405. Combine and map the main aspects of behavioral deviation, the type attributes of the source group, and the type attributes of the potential approaching group.

[0182] S406. Based on a predefined mapping rule base, assign corresponding semantic type labels to the combination. Semantic type labels include social isolation, interest shift, behavioral disorder, or economic pressure.

[0183] S407. Encapsulate the information of semantic type labels, dominant deviation dimensions and their contribution, source group and potential similar groups to generate risk warning signals.

[0184] In step S401, the contribution is obtained by calculating the absolute value of the product of the standardized deviation value of each feature dimension and the corresponding preset dimension weight. This value quantifies the degree of influence of that dimension on the final value of the relative deviation of the group. The preset contribution threshold is determined based on the statistical distribution of the contribution of all students in each dimension in historical data. For example, by analyzing the distribution of contribution values, a critical value that can distinguish between significant abnormal contributions and general fluctuations is selected.

[0185] In step S402, the feature category is a pre-defined category that divides all feature dimensions in the multi-dimensional behavioral feature vector based on behavioral semantics. The spatiotemporal trajectory category includes feature dimensions related to entry and exit locations and time patterns; the consumption pattern category includes feature dimensions related to consumption amount and structural proportions; and the social collaboration category includes feature dimensions related to the frequency of shared activities. Based on the category to which the dominant deviation dimension belongs, it can be determined in which aspect the behavioral deviation mainly occurs.

[0186] In step S403, the target student similarity is calculated using cosine similarity, which is the cosine of the angle between the target student's multidimensional behavioral feature vector and the central feature vector of each behavioral paradigm group. This calculation is performed on the current feature vector extracted at the current analysis time point, as well as the historical feature vector corresponding to historical time points (such as the starting point of adaptive behavior deviation trend).

[0187] In step S404, the source group identifies the group with the highest similarity by comparing the similarity between the target student and each behavioral paradigm group at historical time points. The potential approaching group identifies the group with the largest increase or exceeding the threshold by comparing the increase in the similarity between the target student and each behavioral paradigm group at the current analysis time point compared to historical time points, or by comparing the current similarity with a preset similarity threshold.

[0188] In step S405, the main aspects of behavioral deviation, the type attributes of the source group, and the type attributes of potential approaching groups are combined. The type attributes are a general description of the behavioral patterns of the behavioral paradigm group, such as induction based on the features of its central feature vector. This combination constitutes the input information unit for semantic mapping.

[0189] In step S406, a predefined mapping rule base stores mapping rules formed from domain expert knowledge or historical case summaries. The condition portion of each rule defines a specific combination pattern of the main aspects of behavioral deviation, the source group type attribute, and the potential approaching group type attribute; the conclusion portion specifies the semantic type label corresponding to this combination pattern. The semantic type label is defined based on common risk patterns: social isolation refers to deviations mainly manifested in weakened social collaboration and a tendency towards alienated groups; interest shift refers to deviations mainly manifested in changes in consumption patterns (such as increased entertainment consumption) and a tendency towards specific interest groups; behavioral disorder refers to deviations mainly manifested in disordered spatiotemporal trajectories; and economic pressure refers to deviations mainly manifested in a significant decrease in consumption levels. This step matches the combination generated in step S405 with the conditions in the rule base, assigning the corresponding semantic type label to the current instance.

[0190] In step S407, the semantic type labels, the list of dominant deviation dimensions and their contribution values, and the identification information of the source group and potential similar groups are encapsulated according to a predefined structured format, such as JSON, XML, or a predefined internal data structure. This encapsulation includes fixed fields to carry key information such as semantic type labels, the list of dominant deviation dimensions, contribution values, and the identifiers and type attributes of the source group and potential similar groups. This structured encapsulation ensures that the generated risk warning signal can be accurately and efficiently parsed and used by the subsequent assistance suggestion generation module or other external systems.

[0191] This embodiment identifies the dominant deviation dimension by quantitatively analyzing the contribution of each feature dimension to the relative deviation of the group. It combines this with similarity calculations between the target student and various behavioral paradigm groups at different time points to identify patterns of change in their behavioral paradigm group affiliation. Based on predefined mapping rules, it maps the main aspects of behavioral deviation and changes in group relationships into specific semantic type labels. This embodiment transforms the numerical deviation calculated based on the relative deviation calculation model into a risk warning signal with clear business implications, including specific behavioral attributions and the background of community changes, fundamentally achieving interpretability of the warning results.

[0192] In some embodiments, the main aspects of behavioral deviation, the type attributes of the source group, and the type attributes of potential approaching groups are combined and mapped; based on a predefined mapping rule base, corresponding semantic type labels are assigned to the combination, including:

[0193] The main aspects of behavioral deviation, the type attributes of the source group, and the type attributes of the potential approaching group are encoded into triplet data.

[0194] The triplet data is matched with the rule conditions in the predefined mapping rule base. Each rule in the mapping rule base contains a condition part and a conclusion part. The condition part defines a specific combination pattern of behavioral deviation from the main aspect, source group type attribute and potential approaching group type attribute.

[0195] When the triplet data satisfies part of the condition of a rule, that rule is triggered;

[0196] The conclusion part of the triggered rule is executed. The operations in the conclusion part include directly outputting the predefined semantic type label, or calling the scoring function to calculate the confidence of multiple candidate semantic type labels and selecting the label with the highest confidence to output.

[0197] The output semantic type label is associated with the triple data that triggered the label and the specific rule identifier that was matched, thus completing the combination mapping and label assignment.

[0198] In this embodiment, the triplet data encodes the main aspects of behavioral deviation, the type attributes of the source group, and the type attributes of the potential approaching group into a structured data unit, such as a string array or dictionary with three fields.

[0199] The predefined mapping rule library is stored in a data structure that can be parsed by the program. Each rule contains a condition part and a conclusion part. The condition part defines a specific combination or pattern of values ​​(such as exact value matching, wildcard matching, or belonging to a set) of the behavioral deviation from the main aspect, the source group type attribute, and the potential approaching group type attribute. The conclusion part specifies the action after the condition is met, which can be either directly outputting a preset semantic type label or calling a scoring function to make a decision.

[0200] Matching triplet data with the rule base is achieved by sequentially comparing the values ​​of each field in the triplet with the corresponding patterns defined in the rule conditions section. When all fields of a triplet satisfy the conditions of a certain rule, a successful match is determined, and that rule is triggered.

[0201] If the conclusion of the triggered rule specifies the invocation of a scoring function, then that function is executed. Optionally, the scoring function is implemented based on a historical case library, which stores historical triplet data and their corresponding valid semantic type labels. The function calculates the similarity between the current triplet and each case triplet in the historical case library. Similarity calculation can use field-based weighted cosine similarity or edit distance methods. The function uses the semantic type labels corresponding to the top few cases with the highest similarity as a candidate set, and calculates a comprehensive confidence score for each candidate label. The comprehensive confidence score can be obtained by weighting and summing factors such as the similarity weight of its corresponding case and the overall frequency of the label in historical cases. The function outputs the semantic type label with the highest comprehensive confidence score.

[0202] Alternatively, the scoring function can be implemented based on rule weights and feature matching degrees. Each rule that triggers the scoring function has a conclusion associated with a set of candidate semantic type labels and the base weight of each label. The function calculates a matching degree coefficient based on the degree of match between the values ​​of each field in the current triplet and the rule condition pattern. The base weight of each candidate label is multiplied by the matching degree coefficient to obtain its final score, and the label with the highest score is output.

[0203] After executing the conclusion section, the final determined semantic type label is associated with the original triplet data that triggered this output and the matched rule identifier. This association information is written into the generated risk warning signal, completing the entire process of combination mapping and label allocation.

[0204] This embodiment encodes multi-source information into a standardized triplet data structure and utilizes a structured rule base for precise matching or intelligent scoring, achieving an automated and interpretable mapping from behavioral features and group relationships to specific risk semantic labels. The introduction of a scoring function enhances the system's ability to handle complex and boundary situations, while complete association records ensure the auditability and optimizability of the early warning results.

[0205] In some embodiments, based on risk warning signals, a pre-set intervention strategy library is referenced to generate and output targeted assistance and guidance suggestions, including:

[0206] Analyze risk warning signals and extract semantic type labels, dominant deviation dimension information, and source group and potential similar group information;

[0207] Using semantic type tags as the primary key, a search is performed in a pre-built intervention strategy library, which stores basic intervention strategy templates and historical success case indexes corresponding to different semantic type tags.

[0208] Based on the dominant deviation dimension information, specific intervention action descriptions with the highest correlation to the dominant deviation dimension are selected from the retrieved basic intervention strategy templates.

[0209] Based on information about the source group and potential similar groups, the description of specific intervention actions is parametrically adjusted. The parametric adjustments include suggested intervention personnel, suggested activity scenarios, and suggested communication entry points.

[0210] The parameterized descriptions of specific intervention actions are integrated with similar case summaries matched from the historical success case index to generate a structured draft of assistance and guidance suggestions;

[0211] The draft of the support and guidance recommendations is linked with the basic profile information of the target students, and a summary of key evidence for risk warning signals is attached. The final targeted support and guidance recommendation document is then generated and output.

[0212] In this embodiment, parsing the risk warning signal means reading specific fields in the structured data object, including semantic type label strings, a list of dominant deviation dimensions (containing dimension identifiers and their contribution values), and identifiers and type attribute strings of the source group and potential similar groups.

[0213] The intervention strategy library can be physically represented as a database or configuration file, organized using semantic type tags as primary keys. Each tag is associated with a basic intervention strategy template and a historical success case index. The basic intervention strategy template is a structured document or data object containing general intervention objectives, guiding principles, and multiple optional specific intervention actions for that type of risk. Each action description is accompanied by a tag representing the behavioral dimension it targets. The historical success case index stores identifiers, key feature summaries (such as major deviation dimensions and group changes), and brief descriptions of intervention effects from similar past risk cases.

[0214] When filtering specific intervention action descriptions based on dominant deviation dimension information, the dimension identifiers in the dominant deviation dimension list are matched against the dimension-specific labels attached to each specific intervention action description in the basic intervention strategy template. The matching process considers both exact matching and categorical matching, and a relevance score is calculated for each matched action. This score can be combined with the dimension contribution weight. The specific intervention action descriptions with the highest relevance scores are selected.

[0215] Parametric adjustments are made based on information from the source population and potential similar populations. Specifically, based on the type attributes of these two populations, the variable parts in the descriptions of the selected specific intervention actions are filled in or replaced.

[0216] The process involves retrieving case summaries from a historical success case index that are most similar to current risk warning signals in terms of semantic type and dominant deviation dimension, and integrating them with a parametrically adjusted description of specific intervention actions. The integration method involves embedding the case summaries as "reference examples" or "experience-based evidence" modules into a draft of support and guidance recommendations centered on intervention actions, forming a structured draft text that includes recommended actions, theoretical basis, and reference cases.

[0217] The draft support and guidance recommendations are linked to basic student profile information (such as name, grade, and major) obtained from the student basic information database, and key evidence summaries extracted from risk warning signals (such as semantic type tags and core deviation dimensions) are attached. These contents are then combined according to a predefined document template to generate a complete and highly targeted support and guidance recommendation document, which can be directly reviewed and used by counselors or psychological counselors.

[0218] This embodiment achieves automated generation of specific, personalized, and actionable assistance suggestions from risk warning by analyzing early warning signals, retrieving strategy libraries, filtering and adjusting intervention actions based on multi-dimensional information, and integrating historical cases. This ensures a close correlation between intervention suggestions and risk causes, incorporates successful historical experiences, improves the pertinence and effectiveness of assistance measures, and forms a complete closed loop from risk identification to intervention suggestions.

[0219] Please see Figure 5 In a second aspect, this embodiment also provides a student psychological assessment and early warning system 1 based on group behavior clustering analysis, applicable to the method described in the first aspect. The system includes a data acquisition and fusion module 11, a dynamic group clustering module 12, a group baseline calculation module 13, an individual deviation calculation module 14, a persistent trend analysis module 15, a risk signal generation module 16, and a support suggestion generation module 17. The data acquisition and fusion module 11 is configured to acquire and integrate multi-source campus behavior time-series data of the target student group; the dynamic group clustering module 12 is configured to extract group behavior feature vectors based on multi-source campus behavior time-series data and dynamically divide multiple behavior paradigm groups using an unsupervised clustering algorithm; the group baseline calculation module 13 is configured to calculate the multi-dimensional behavior baseline of each behavior paradigm group, the multi-dimensional behavior baseline representing the behavior paradigm group within a preset time period. The system includes several modules: a normal behavior pattern range; an individual deviation calculation module 14, configured to calculate the relative deviation between the target student's current behavioral feature vector and the multidimensional behavioral baseline of their historical behavioral paradigm group; a persistent trend analysis module 15, configured to perform persistent trend analysis on the relative deviation of the group based on a sliding time window, determining adaptive behavioral deviation when a statistically significant persistent increasing trend in the relative deviation of the group is detected; a risk signal generation module 16, configured to perform deviation dimension attribution analysis on target students who have experienced adaptive behavioral deviation, identify the specific behavioral dimensions that dominate the deviation, and generate a semantically typed risk warning signal based on the changing patterns of their behavioral paradigm group affiliation; and a support suggestion generation module 17, configured to generate and output targeted support and guidance suggestions based on the risk warning signal, in conjunction with a pre-set intervention strategy library.

[0220] In this embodiment, each module of the system can be implemented through software programming and deployed on a server or cloud platform. The data acquisition and fusion module 11 obtains raw data by calling the API interfaces of various campus information systems and performs cleaning, alignment, and aggregation. The dynamic group clustering module 12 executes feature extraction and clustering algorithms, outputting a dynamically updated list of behavioral paradigm groups and their central feature vectors. The group baseline calculation module 13 calculates the statistical baseline interval for each behavioral paradigm group based on historical feature vector sequences. The individual deviation calculation module 14 receives the current student's feature vector and calculates the relative deviation of the group by querying its historical group baseline. The continuous trend analysis module 15 applies a statistical trend detection algorithm to the deviation time series. The risk signal generation module 16 performs contribution analysis and rule matching, outputting a structured early warning signal. The support suggestion generation module 17 parses the signal, retrieves the intervention strategy library, and generates a final suggestion document. The modules of this system collaborate sequentially, achieving full-process automation from automatic collection of multi-source heterogeneous data to intelligent generation of targeted support suggestions, providing an efficient, accurate, and operable decision support tool for campus mental health management.

[0221] By adopting the above technical solutions, this invention differs from existing technologies and has the following beneficial effects: By dynamically dividing behavioral paradigm groups and establishing their multidimensional behavioral baselines, the risk assessment standard is transformed from an absolute threshold to a normal range relative to the individual's community, effectively overcoming the false alarm problem caused by the natural differences in students' behavioral patterns; by calculating the relative deviation of the group and conducting continuous trend analysis based on a sliding time window, early signals of continuous deterioration of behavioral patterns can be captured, achieving forward-looking early warning; furthermore, by attribution analysis of the deviation dimension and combining it with the changing patterns of the affiliation of behavioral paradigm groups, a semantic risk warning signal with clear business meaning is generated, transforming numerical deviation into an interpretable result containing specific behavioral attributions and the background of community changes; finally, based on the semantic warning signal, an intervention strategy library is associated to generate and output targeted assistance and guidance suggestions, forming a complete data-driven closed loop from risk identification to intervention suggestions, improving the practicality and operability of campus psychological risk early warning.

[0222] Finally, it should be noted that although the above embodiments have been described in the description and drawings of this invention, this should not limit the scope of patent protection of this invention. Any technical solutions that are based on the essential concept of this invention, utilize the content described in the description and drawings of this invention to make equivalent structural or procedural substitutions or modifications, as well as the direct or indirect application of the technical solutions of the above embodiments to other related technical fields, are all included within the scope of patent protection of this invention.

Claims

1. A student psychological assessment and early warning method based on group behavior clustering analysis, characterized in that, include: Acquire multi-source time-series data on campus behavior of the target student group; Based on the multi-source campus behavior time-series data, group behavior feature vectors are extracted, and multiple behavioral paradigm groups are dynamically divided using an unsupervised clustering algorithm. Calculate the multidimensional behavioral baseline for each behavioral paradigm group, wherein the multidimensional behavioral baseline represents the range of normal behavioral patterns of the behavioral paradigm group within a preset time period; Calculate the relative deviation between the target student’s current behavioral feature vector and the multidimensional behavioral baseline of the behavioral paradigm group to which it historically belongs; Based on a sliding time window, a continuous trend analysis is performed on the relative deviation of the group. When a statistically significant continuous increasing trend is detected in the relative deviation of the group, it is determined that an adaptive behavior deviation has occurred. For target students who exhibit adaptive behavioral deviations, conduct deviation dimension attribution analysis to identify the specific behavioral dimensions that dominate the deviations, and generate risk warning signals with semantic types by combining the changing patterns of their behavioral paradigm group affiliation. Based on the aforementioned risk warning signals, a pre-set intervention strategy library is referenced to generate and output targeted assistance and guidance suggestions.

2. The student psychological assessment and early warning method based on group behavior clustering analysis according to claim 1, characterized in that, Based on the aforementioned multi-source campus behavior time-series data, group behavior feature vectors are extracted, and an unsupervised clustering algorithm is used to dynamically divide multiple behavioral paradigm groups, including: From the multi-source campus behavior time-series data, spatiotemporal trajectory features, consumption pattern features, and social collaboration features are extracted according to a preset period. The spatiotemporal trajectory features include the time distribution of regular entry and exit from core teaching and living places. The consumption pattern features include the consumption level and consumption structure ratio. The social collaboration features are calculated based on the frequency of common occurrence at the same spatiotemporal point. The spatiotemporal trajectory features, consumption pattern features, and social collaboration features are standardized to eliminate differences in units, and weighted based on the information entropy of the features or predefined importance weights. The weighted feature values ​​of each dimension are then concatenated to construct a unified multidimensional behavioral feature vector for each student. The multidimensional behavioral feature vector set is used as input, and unsupervised learning is performed using a density-based clustering algorithm. The density-based clustering algorithm identifies core points, boundary points, and noise points by setting a neighborhood radius and a minimum number of points threshold, and connects mutually reachable core points and boundary points in their neighborhoods to form multiple density-connected sample sets. Each density-connected set of samples is defined as a behavioral paradigm group, and the central feature vector of each behavioral paradigm group is calculated. The process of feature extraction, weighted fusion, and density-based clustering is periodically repeated to update the membership composition and central feature vector of the behavioral paradigm group, thereby achieving dynamic partitioning of the behavioral paradigm group.

3. The student psychological assessment and early warning method based on group behavior clustering analysis according to claim 1, characterized in that, Calculate the multidimensional behavioral baseline for each of the aforementioned behavioral paradigm groups, including: For each behavioral paradigm group, obtain a multidimensional behavioral feature vector sequence of all its members within a preset historical time period; For each feature dimension in the multidimensional behavioral feature vector sequence, calculate the statistical distribution of each feature dimension, whereby the statistical distribution includes the mean and variance. Based on the statistical distribution of each feature dimension, a multidimensional behavioral baseline interval is constructed. The multidimensional behavioral baseline interval is jointly defined by the normal value range of each feature dimension. The normal value range is determined by the confidence interval formed by the mean plus or minus a certain number of variances. The multidimensional behavioral baseline interval is associated with and stored with the central feature vector of the behavioral paradigm group, serving as a quantitative reference benchmark to characterize the normal behavioral pattern of the behavioral paradigm group.

4. The student psychological assessment and early warning method based on group behavior clustering analysis according to claim 1, characterized in that, Calculate the relative deviation between the target student's current behavioral feature vector and the multidimensional behavioral baseline of their historical behavioral paradigm group, including: Identify the behavioral paradigm group to which the target student belonged with the highest intensity before the current analysis point, and use this as the historical behavioral paradigm group to which the student belonged. Obtain the multidimensional behavioral baseline interval and its central feature vector of the historical behavioral paradigm group; The original deviation vector is obtained by calculating the difference between the multidimensional behavioral feature vector of the target student at the current analysis time point and the central feature vector in each dimension. For each dimension deviation value in the original deviation vector, normalization is performed using the normal value range of the corresponding dimension in the multidimensional behavior baseline interval to obtain the standardized deviation value. The standardized deviation values ​​are weighted and summed according to preset dimension weights to calculate the Mahalanobis distance, which is used as the initial deviation. The initial deviation is compared with a preset deviation threshold. If the initial deviation exceeds the deviation threshold, it is determined that the target student has exceeded the normal behavior pattern range of the historical behavioral paradigm group, and the initial deviation is used as the output value of the group's relative deviation. If the initial deviation does not exceed the deviation threshold, the relative deviation of the group is output as zero.

5. The student psychological assessment and early warning method based on group behavior clustering analysis according to claim 4, characterized in that, For each dimension deviation value in the original deviation vector, normalization is performed using the normal value range of the corresponding dimension in the multidimensional behavioral baseline interval to obtain the standardized deviation value, including: For any dimension deviation value in the original deviation vector, obtain the upper limit and lower limit of the normal value range of the corresponding dimension in the multidimensional behavior baseline interval; Calculate the difference between the upper limit of the normal value range and the lower limit of the normal value range, and use it as the length of the normal value range for this dimension; Determine the sign of the dimensional deviation value. If the dimensional deviation value is positive, it indicates that the behavioral feature value of the target student in that dimension is higher than the corresponding value of the central feature vector. If the dimension deviation value is positive, then the dimension deviation value is divided by the difference between the upper limit of the normal value range and the corresponding value of the central feature vector to obtain the positive standardized deviation value of the dimension. If the dimensional deviation value is negative, it indicates that the behavioral feature value of the target student in that dimension is lower than the corresponding value of the central feature vector; If the dimension deviation value is negative, then the absolute value of the dimension deviation value is divided by the difference between the corresponding value of the central feature vector and the lower limit of the normal value range to obtain the negative standardized deviation value of the dimension. The positive or negative standardized deviation value is output as the standardized deviation value for that dimension.

6. The student psychological assessment and early warning method based on group behavior clustering analysis according to claim 1, characterized in that, Based on a sliding time window, a continuous trend analysis of the relative deviation of the group is performed, including: Using the current analysis time point as a baseline, a sliding time window is formed by extracting a sequence of relative deviations of the population that includes multiple consecutive time points. For the population relative deviation sequence within the sliding time window, a trend line of change of the population relative deviation sequence is fitted using linear regression, and the slope of the trend line is calculated. For the population relative deviation sequence within the sliding time window, the cumulative sum control algorithm is applied to calculate the cumulative sum statistic at each time point. The cumulative sum statistic reflects the cumulative offset of the sequence value relative to the historical mean. Set the trend slope threshold and cumulative and statistical control limits for the trend line; If the slope of the trend line is greater than the trend slope threshold, and the cumulative sum statistic continues to exceed the cumulative sum statistic control limit, then it is determined that the relative deviation of the group shows a statistically significant and continuous increasing trend. Record the starting point and current intensity of the sustained increasing trend as temporal evidence for determining the occurrence of adaptive behavioral deviation.

7. The student psychological assessment and early warning method based on group behavior clustering analysis according to claim 1, characterized in that, For target students exhibiting adaptive behavioral deviations, attribution analysis of deviation dimensions is conducted to identify the specific behavioral dimensions that dominate the deviation. Combined with changes in their behavioral paradigm group affiliation patterns, semantically typed risk warning signals are generated, including: The feature dimension whose contribution exceeds a preset contribution threshold when calculating the relative deviation of the group is extracted and used as the dominant deviation dimension; Based on the feature category to which the dominant deviation dimension belongs, the main aspects of behavioral deviation are determined, and the feature categories include spatiotemporal trajectory, consumption patterns, and social collaboration. Calculate the similarity between the target student and the central feature vector of each behavioral paradigm group at the current analysis time point and at historical time points, and denote it as the target student similarity. Identify the behavioral paradigm group with the highest historical affiliation strength of the target students as the source group, and identify the behavioral paradigm group with the fastest current similarity growth to the target students or that has exceeded the preset similarity threshold as the potential approaching group; The main aspects of the behavioral deviation, the type attributes of the source group, and the type attributes of the potential approaching group are combined and mapped. Based on a predefined mapping rule library, corresponding semantic type labels are assigned to combinations, including social isolation, interest shift, behavioral disorder, or economic pressure. The semantic type label, the dominant deviation dimension and its contribution, and the information of the source group and the potential similar group are encapsulated to generate the risk warning signal.

8. The student psychological assessment and early warning method based on group behavior clustering analysis according to claim 7, characterized in that, The main aspects of the behavioral deviation, the type attributes of the source group, and the type attributes of the potential similar groups are combined and mapped; based on a predefined mapping rule library, corresponding semantic type labels are assigned to the combination, including: The main aspects of the behavioral deviation, the type attributes of the source group, and the type attributes of the potential approaching group are encoded into triplet data. The triplet data is matched with the rule conditions in the predefined mapping rule base. Each rule in the mapping rule base includes a condition part and a conclusion part. The condition part defines a specific combination pattern of behavioral deviation from the main aspect, source group type attribute and potential approaching group type attribute. When the triplet data satisfies part of the condition of a certain rule, the rule is triggered; The conclusion part of the triggered rule is executed. The operations of the conclusion part include directly outputting the predefined semantic type label, or calling the scoring function to calculate the confidence of multiple candidate semantic type labels and selecting the label with the highest confidence to output. The output semantic type label is associated with the triple data that triggered the label and the specific rule identifier that was matched, thus completing the combination mapping and label assignment.

9. The student psychological assessment and early warning method based on group behavior clustering analysis according to claim 1, characterized in that, Based on the aforementioned risk warning signal, and in conjunction with a pre-set intervention strategy library, targeted assistance and guidance suggestions are generated and output, including: The risk warning signal is analyzed to extract semantic type labels, dominant deviation dimension information, and source group and potential similar group information; Using the semantic type tag as the primary key, a search is performed in a pre-set intervention strategy library, which stores basic intervention strategy templates and historical success case indexes corresponding to different semantic type tags. Based on the dominant deviation dimension information, specific intervention action descriptions with the highest correlation to the dominant deviation dimension are selected from the retrieved basic intervention strategy templates. Based on the information of the source group and potential similar groups, the description of the specific intervention action is parametrically adjusted, including the suggested intervention personnel, suggested activity scenarios, and suggested communication entry points; The parameterized descriptions of specific intervention actions are integrated with similar case summaries matched from the historical success case index to generate a structured draft of assistance and guidance suggestions; The draft assistance and guidance recommendations are linked with the basic profile information of the target students, and a summary of key evidence for the risk warning signals is attached. The final targeted assistance and guidance recommendation document is then generated and output.

10. A student psychological assessment and early warning system based on group behavior clustering analysis, characterized in that, The system applicable to the method of any one of claims 1 to 9 comprises: The data acquisition and fusion module is configured to acquire and integrate multi-source time-series campus behavior data of the target student group; The dynamic group clustering module is configured to extract group behavior feature vectors based on the multi-source campus behavior time series data, and dynamically divide multiple behavior paradigm groups using an unsupervised clustering algorithm; The group baseline calculation module is configured to calculate the multidimensional behavioral baseline for each of the behavioral paradigm groups, wherein the multidimensional behavioral baseline represents the range of normal behavioral patterns of the behavioral paradigm group within a preset time period; The individual deviation calculation module is configured to calculate the relative deviation between the target student’s current behavioral feature vector and the multidimensional behavioral baseline of the behavioral paradigm group to which it historically belongs; The persistent trend analysis module is configured to perform persistent trend analysis on the relative deviation of the group based on a sliding time window. When a statistically significant persistent increasing trend in the relative deviation of the group is detected, it is determined that an adaptive behavior deviation has occurred. The risk signal generation module is configured to perform attribution analysis on the deviation dimensions of target students who have undergone adaptive behavioral deviations, identify the specific behavioral dimensions that dominate the deviations, and generate risk warning signals with semantic types by combining the changing patterns of their behavioral paradigm group affiliation. The assistance suggestion generation module is configured to generate and output targeted assistance and guidance suggestions based on the risk warning signal and a pre-set intervention strategy library.

Images