Student psychological data analysis and early warning system based on artificial intelligence

Abstract

The invention relates to the technical field of health information processing, and particularly discloses a student psychological data analysis and early warning system based on artificial intelligence, which is provided with a data set complexity judgment module, a data slice judgment module, a GPU compatible matching module and a psychological data analysis and early warning module. The psychological health monitoring platform evaluates the comprehensive complexity by collecting a multi-source psychological data set of a student, and determines the number of psychological data slices and the respective complexity according to the comprehensive complexity. The platform monitors execution data of each GPU in real time, and evaluates an execution quality initial index and compatible complexity of the GPU. Then, through data processing, the comprehensive complexity of the psychological data slices is matched with the compatible complexity of each GPU, and distribution of the psychological data slices and the GPUs is completed. And finally, the platform analyzes the psychological data of the students by using the distribution relation, and finally realizes analysis and early warning of the psychological data of the students.

Description

Technical Field

[0001] The present invention relates to the technical field of health information processing, and particularly to a student psychological data analysis and early warning system based on artificial intelligence. Background Art

[0002] The wide application of artificial intelligence technology in various fields has achieved remarkable results, providing new ideas and methods for solving the problems of student psychological data analysis and early warning. Artificial intelligence has powerful data processing capabilities, pattern recognition capabilities, and predictive analysis capabilities, which can deeply mine and analyze a large amount of student psychological data, discover potential signs of psychological problems, issue early warnings in a timely manner, and provide strong support for mental health educators.

[0003] For example, the invention patent with the publication number CN118841178B discloses a method and system for evaluating students' psychological qualities, which relates to the technical field of health data analysis. The entire evaluation system is divided into four parts, namely a health data summary module, a health evaluation module, a data storage and analysis module, and a statistical adjustment module. The technical key points are as follows: collecting corresponding data according to different types of students can more comprehensively understand and support the mental health needs of students of different ages. Based on multiple factors reflecting students' mental health in the corresponding data and through comprehensive evaluation by analysis and calculation, it can more comprehensively capture multi-dimensional information of students' mental health. When triggering the psychological counseling mechanism subsequently, it helps to improve the response efficiency and effect of students' mental health problems. According to the proportion value of the number of students triggering the psychological counseling mechanism to the total number of students in the school, the employment quantity of psychological teachers is dynamically adjusted, realizing the effective allocation and utilization of resources.

[0004] For example, the invention patent with the publication number CN119361142A discloses a mental health assessment and early warning analysis system based on health big data, including a data collection module, a monitoring module, an analysis and determination module, an information management module, and an early warning module. The monitoring module monitors the mental health status of all students in real time, and the analysis and determination module analyzes the monitoring data, so as to determine whether a student has mental health problems and issue an early warning in a timely manner through the early warning module when a student has mental health problems.

[0005] Combined with the above technical solutions, it is found that the monitoring data of students' mental health is more multi-source and large in quantity. The existing technical solutions for monitoring students' mental health directly use the obtained data without targeted processing, which easily brings unnecessary processing increments to data processing devices, reduces the processing speed of the devices, and at the same time there is a problem of inaccurate psychological data analysis due to insufficient flexibility of device operation. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides a student psychological data analysis and early warning system based on artificial intelligence, which can effectively solve the problems involved in the above-mentioned background technology.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: A student psychological data analysis and early warning system based on artificial intelligence includes a dataset complexity determination module, which is used to extract a multi-source psychological dataset through a mental health monitoring platform and comprehensively determine the comprehensive complexity of the multi-source psychological dataset; a data slice determination module, which is used to match the number of psychological data slices according to the comprehensive complexity of the multi-source psychological dataset, and finally extract the comprehensive complexity of each psychological data slice; a GPU compatible matching module, which is used to collect the execution data of each GPU of the mental health monitoring platform in real time, determine the initial index of the execution quality of each GPU, and match the compatible complexity of each GPU; a psychological data analysis and early warning module, which is used to perform data processing on the comprehensive complexity of each psychological data slice and the compatible complexity of each GPU, complete the allocation of psychological data slices and GPUs, and finally the mental health monitoring platform analyzes and warns the student psychological dataset.

[0008] As a further solution, comprehensively determining the comprehensive complexity of the multi-source psychological dataset specifically includes: The multi-source psychological dataset includes each student's psychological dataset. Determine the sub-complexity of each student's psychological dataset and perform mean processing to finally obtain the comprehensive complexity of the multi-source psychological dataset; Extract feature sets from each student's psychological dataset, specifically including the total data volume of each student's psychological dataset, the number of data categories of each student's psychological dataset, the data time span of each student's psychological dataset, and the information entropy of each student's psychological dataset; Perform weighted aggregation and normalization processing on the total data volume of each student's psychological dataset, the number of data categories of each student's psychological dataset, the time span of each student's psychological dataset, and the information entropy of each student's psychological dataset in sequence to obtain the sub-complexity of each student's psychological dataset.

[0009] As a further solution, the process of extracting the comprehensive complexity of each psychological data slice is specifically as follows: Randomly divide the multi-source psychological dataset of each student according to the number of psychological data slices to obtain each psychological data slice; Perform mean processing on the sub-complexity of each student's psychological dataset included in each psychological data slice to obtain the comprehensive complexity of each psychological data slice.

[0010] As a further solution, determine the initial execution quality indicators for each GPU. The specific determination process is as follows: The execution data of each GPU specifically includes the number of floating-point operations of each GPU within the quality determination period, the instruction execution efficiency of each GPU within the quality determination period, the video memory bandwidth of each GPU within the quality determination period, and the data throughput of each GPU within the quality determination period. Extract the defined number of floating-point operations, the defined instruction execution efficiency, the defined video memory bandwidth, and the defined data throughput from the mental health database. Perform weighted aggregation processing on the deviation degrees between the number of floating-point operations of each GPU within the quality determination period and the defined number of floating-point operations, the deviation degrees between the instruction execution efficiency of each GPU within the quality determination period and the defined instruction execution efficiency, the deviation degrees between the video memory bandwidth of each GPU within the quality determination period and the defined video memory bandwidth, and the deviation degrees between the data throughput of each GPU within the quality determination period and the defined data throughput in sequence to obtain the initial execution quality indicators for each GPU.

[0011] As a further solution, complete the slicing of psychological data and the allocation to GPUs. The specific analysis process is as follows: Perform difference processing on the comprehensive complexity of each psychological data slice and the compatible complexity of each GPU respectively to obtain the complexity deviation values between each GPU and each psychological data slice. Denote the GPU corresponding to the minimum complexity deviation value with a certain psychological data slice as the designated GPU for this psychological data slice. Traverse the complexity deviation values between all GPUs and psychological data slices in sequence to obtain the designated GPUs for each psychological data slice, and complete the slicing of psychological data and the allocation to GPUs.

[0012] As a further solution, analyze and give early warnings for the student psychological data set. The specific analysis process is as follows: Allocate each psychological data slice to the corresponding designated GPU for data analysis, and collect the total processing delay duration of the mental health monitoring platform. Compare it with the predefined total defined processing delay duration to obtain a comparison result. The mental health monitoring platform determines whether to adjust the processing efficiency of the GPU based on the comparison result. Finally, the mental health monitoring platform analyzes the psychological state of the students, visualizes the analysis results for early warning, and outputs the early warning results to each receiving terminal to which it belongs.

[0013] As a further solution, the processing efficiency of the GPU is adjusted. The specific adjustment process is: the total processing delay duration of the mental health monitoring platform is subtracted from the total processing delay definition duration to obtain a delay duration deviation value, and the clock frequency adjustment value is matched to adjust the processing efficiency of each GPU. At the same time, the execution quality real-time indicator of each GPU is monitored in real time and compared with the predefined execution quality threshold. If the execution quality real-time indicator of a GPU is less than the execution quality threshold, the core frequency adjustment value of the GPU is matched according to the difference between the execution quality real-time indicator of the GPU and the execution quality threshold, which is used to reduce the core frequency of the GPU.

[0014] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0015] (1) The present invention provides a student psychological data analysis and early warning system based on artificial intelligence. The psychological health monitoring platform collects multi-source psychological data sets of students, evaluates their comprehensive complexity, and determines the number of psychological data slices and their respective complexities accordingly. The platform monitors the execution data of each GPU in real time, evaluates its initial execution quality indicators and compatible complexity. Subsequently, through data processing, the comprehensive complexity of the psychological data slices is matched with the compatible complexity of each GPU, and the allocation of psychological data slices and GPUs is completed. Finally, the platform uses this allocation relationship to analyze the students' psychological data, and finally realizes the analysis and early warning of the students' psychological data.

[0016] (2) The present invention extracts multi-source psychological data sets of each student and comprehensively determines the comprehensive complexity of the multi-source psychological data sets. It can measure the complexity of the data as a whole, comprehensively consider factors such as the total data volume, the number of data categories, the time span, and information entropy, and help the system understand the characteristics of the entire data set. It is conducive to the rational planning of data processing procedures, such as determining whether a more efficient data processing algorithm needs to be adopted according to the complexity, or allocating more computing resources in advance, thereby laying the foundation for the subsequent accurate analysis of student psychological data.

[0017] (3) The present invention can refine the analysis of data by extracting the comprehensive complexity of each psychological data slice. After dividing the data set into slices, by calculating the complexity of each slice, the complexity differences of different parts of the data can be more accurately understood. In this way, when processing data, computing resources can be allocated in a targeted manner according to the complexity of each slice, and complex slices can be allocated to GPUs with stronger performance for processing, thereby improving processing efficiency and avoiding resource waste, thereby ensuring that the analysis of students' psychological data is more accurate and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention will be further described with reference to the accompanying drawings. However, the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the following drawings without creative efforts.

[0019] Figure 1 It is a schematic diagram of the system module connection of the present invention.

[0020] Figure 2 It is a schematic diagram of each receiving terminal belonging to the mental health monitoring platform. Specific embodiments

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0022] Refer to Figure 1 As shown, the embodiments of the present invention provide a technical solution: a student psychological data analysis and early warning system based on artificial intelligence, including a dataset complexity determination module, a data slicing determination module, a GPU compatible matching module, and a psychological data analysis and early warning module.

[0023] The embodiments of the present invention provide a technical solution: a student psychological data analysis and early warning system based on artificial intelligence, further including: a mental health database for storing the preset values of the floating-point operation definition times, instruction execution definition efficiency, video memory definition bandwidth, definition data throughput, and various factors.

[0024] The dataset complexity determination module is connected to the data slicing determination module, the data slicing determination module is connected to the GPU compatible matching module, the GPU compatible matching module is connected to the psychological data analysis and early warning module, and the dataset complexity determination module, the data slicing determination module, the GPU compatible matching module, and the psychological data analysis and early warning module are all connected to the mental health database.

[0025] The dataset complexity determination module is used to extract multi-source psychological datasets through the mental health monitoring platform and comprehensively determine the comprehensive complexity of the multi-source psychological datasets.

[0026] The above-mentioned mental health monitoring platform is specifically an intelligent system (such as the Xiaozhi model) that integrates psychological evaluation, crisis warning, and intervention support based on big data and artificial intelligence technologies to meet the mental health needs of students in the K12 education stage. Its core functions include: integrating multi-source data from online (student message interaction, mood records, tree hole information) and offline (behavioral data, psychological evaluation results), that is, each student's psychological data set, to comprehensively capture the mental state of students; identifying psychological crisis signals such as anxiety and depression through algorithms, triggering warnings in combination with preset rules, and providing intervention suggestions; generating personalized mental health reports for students, providing instant psychological counseling relying on AI digital human technology (such as "Xiaozhi Tongxue"), and at the same time providing data-based references for school psychological counseling rooms. Compared with the traditional model, this platform realizes the dynamic monitoring and precise intervention of students' mental health by reducing the stigma, improving the convenience of intervention and the accuracy of evaluation.

[0027] Specifically, comprehensively determine the comprehensive complexity of the multi-source psychological data set, which specifically includes:

[0028] The multi-source psychological data set includes each student's psychological data set. Determine the sub-complexity of each student's psychological data set and perform mean processing to finally obtain the comprehensive complexity of the multi-source psychological data set.

[0029] Extract the feature set from each student's psychological data set, which specifically includes the total data volume of each student's psychological data set, the number of data categories in each student's psychological data set, the data time span of each student's psychological data set, and the information entropy of each student's psychological data set.

[0030] The above-mentioned information entropy can be specifically calculated through the existing Shannon entropy formula, which is used to measure the uncertainty of the data distribution in each student's psychological data set. The higher the entropy value, the greater the degree of data dispersion and the higher the complexity.

[0031] Perform weighted aggregation and normalization processing on the total data volume of each student's psychological data set, the number of data categories in each student's psychological data set, the time span of each student's psychological data set, and the information entropy of each student's psychological data set in turn to obtain the sub-complexity of each student's psychological data set. The specific analysis method is as follows:

[0032] In the formula, is the sub-complexity of the i-th student's psychological data set, where i is the number of each student's psychological data set, , M is the total amount of the student's psychological data set, is the total data volume of the i-th student's psychological data set, is the number of data categories in the i-th student's psychological data set, is the time span of the i-th student's psychological data set, is the information entropy of the i-th student psychological dataset, is the weight factor corresponding to the predefined total data volume in the mental health database, is the weight factor corresponding to the predefined number of data categories in the mental health database, is the weight factor corresponding to the predefined time span in the mental health database, is the weight factor corresponding to the predefined information entropy in the mental health database.

[0033] It should be noted that the sub-complexity of each of the above-mentioned student psychological datasets, expressed as the individual complexity corresponding to each student psychological dataset, is used to obtain the comprehensive complexity of the multi-source psychological datasets through mean processing.

[0034] It should be noted that the above-mentioned total data volume refers to the total number of records contained in a single student psychological dataset, reflecting the data scale; the number of data categories refers to the number of different data types contained in the dataset, such as structured assessment scores, semi-structured text feedback, unstructured audio and video records, etc.; the time span refers to the time range of data collection, for example, the range from enrollment to the current time, reflecting the timeliness and continuity of the data.

[0035] Among them, the weight factor corresponding to the total data volume, the weight factor corresponding to the number of data categories, the weight factor corresponding to the time span, and the weight factor corresponding to the information entropy are all obtained by extracting from the mental health database, and the mapping relationship therein can be one-to-one or one-to-many. For example, the total data volume, the number of data categories, the time span, and the information entropy respectively form a mapping set with the weight factor corresponding to the predefined total data volume, the weight factor corresponding to the number of data categories, the weight factor corresponding to the time span, and the weight factor corresponding to the information entropy in the mental health database. Substituting the real-time total data volume, the number of data categories, the time span, and the information entropy into the mapping set, the weight factor corresponding to the total data volume, the weight factor corresponding to the number of data categories, the weight factor corresponding to the time span, and the weight factor corresponding to the information entropy are obtained.

[0036] In this embodiment, through the multivariate analysis of the total data volume, the number of data categories, the time span, and the information entropy, the correlation between these parameters is considered. The larger the total data volume, the larger the data scale is reflected, and usually the more abundant the data types covered (such as including behavior logs, text feedback, and audio and video records at the same time), resulting in an increase in the information entropy, so that the sub-complexity of the student psychological dataset is greatly increased; when the time span is larger, long-term data (such as a psychological file covering 3 years) may be more dispersed due to changes in the student's mental state, resulting in an increase in the information entropy value, which has a negative impact on the sub-complexity of the student psychological dataset.

[0037] The data slicing determination module is used to match the number of psychological data slices according to the comprehensive complexity of the multi-source psychological data set, and finally extract the comprehensive complexity of each psychological data slice.

[0038] Further, the process of matching the number of psychological data slices is as follows:

[0039] Match the comprehensive complexity of the multi-source psychological data set with the number of psychological data slices corresponding to each predefined comprehensive complexity interval, determine the interval to which the comprehensive complexity of the multi-source psychological data set belongs, obtain the number of psychological data slices corresponding to this interval, and use it as the psychological data slices of the psychological data set.

[0040] Specifically, the process of extracting the comprehensive complexity of each psychological data slice is as follows:

[0041] Randomly divide the multi-source psychological data set of each student according to the number of psychological data slices to obtain each psychological data slice.

[0042] The above-mentioned psychological data slices specifically refer to the sub-data sets formed by randomly dividing the multi-source psychological data sets of multiple students. Each slice contains all the records of the original data of multiple students, and is used to disperse the processing load and match the GPU resources.

[0043] Random division is used to avoid large deviations in data characteristics between slices, ensure that the comprehensive complexity of each slice is closer to the average value of the overall data set. Reduce the drastic change in slice complexity caused by time span or local concentration of information entropy. Slices with a uniform distribution are more likely to match the compatible complexity of the GPU, avoiding resource congestion caused by a slice with too high complexity. At the same time, according to the slice-level granularity, resource scheduling can be executed more clearly. For example, allocate high-complexity slices to high-performance GPUs and low-complexity slices to conventional GPUs to maximize the cluster efficiency.

[0044] Perform mean processing on the sub-complexities of the psychological data sets of each student included in each psychological data slice to obtain the comprehensive complexity of each psychological data slice.

[0045] The GPU compatible matching module is used to collect the execution data of each GPU in the mental health monitoring platform in real time, determine the initial index of the execution quality of each GPU, and match the compatible complexity of each GPU.

[0046] Further, the process of determining the initial index of the execution quality of each GPU is as follows:

[0047] The execution data of each GPU, specifically including the number of floating-point operations of each GPU within the quality determination cycle, the instruction execution efficiency of each GPU within the quality determination cycle, the video memory bandwidth of each GPU within the quality determination cycle, and the data throughput of each GPU within the quality determination cycle.

[0048] The above-mentioned execution data of each GPU can be extracted from the execution reports of each GPU. The quality determination cycle is a period of time used to determine the execution quality of the GPU.

[0049] Extract the defined number of floating-point operations, defined instruction execution efficiency, defined video memory bandwidth, and defined data throughput from the mental health database.

[0050] The deviation degrees between the number of floating-point operations of each GPU within the quality determination cycle and the defined number of floating-point operations, the deviation degrees between the instruction execution efficiency of each GPU within the quality determination cycle and the defined instruction execution efficiency, the deviation degrees between the video memory bandwidth of each GPU within the quality determination cycle and the defined video memory bandwidth, and the deviation degrees between the data throughput of each GPU within the quality determination cycle and the defined data throughput are weighted and aggregated in sequence to obtain the initial execution quality indicators of each GPU. The specific analysis method is as follows:

[0051] In the formula, is the initial execution quality indicator of the j-th GPU, where j is the number of each GPU, , D is the total number of GPUs, is the number of floating-point operations of the j-th GPU within the quality determination cycle, is the defined number of floating-point operations, is the instruction execution efficiency of the j-th GPU within the quality determination cycle, is the defined instruction execution efficiency, is the video memory bandwidth of the j-th GPU within the quality determination cycle, is the defined video memory bandwidth, is the defined data throughput, is the data throughput of the j-th GPU within the quality determination cycle, is the weight factor corresponding to the defined number of floating-point operations predefined in the mental health database, is the weight factor corresponding to the defined instruction execution efficiency predefined in the mental health database, is the weight factor corresponding to the defined video memory bandwidth predefined in the mental health database, is the weight factor corresponding to the defined data throughput predefined in the mental health database.

[0052] The initial execution quality indicators of the above GPUs represent the performance and functions of the GPUs measured by a series of key parameters such as the number of floating-point operations and the memory bandwidth. The initial execution quality indicators directly affect the efficiency, accuracy, and stability of the GPUs when executing tasks.

[0053] It should be explained that the above instruction execution efficiency refers to the number of instructions that the GPU processor can execute per second, which is one of the important indicators for measuring the performance of the GPU processor, and the unit is millions of instructions per second.

[0054] It should be explained that the above number of floating-point operations refers to the total number of floating-point operations completed by the GPU within the quality determination cycle, reflecting the processing ability of compute-intensive tasks (such as matrix multiplication and convolution operations); the instruction execution efficiency refers to the effective degree of the GPU in executing instructions, and a high value means less instruction waiting time and resource waste; the memory bandwidth refers to the data transfer rate (GB / s) between the GPU memory and the core, which directly affects the read / write efficiency of large-scale data; the data throughput refers to the number of tasks that the GPU can process per unit time. The higher the data throughput, the stronger the processing ability of the GPU.

[0055] Among them, the weight factors corresponding to the number of floating-point operations, the weight factors corresponding to the instruction execution efficiency, the weight factors corresponding to the memory bandwidth, and the weight factors corresponding to the data throughput are all obtained by extracting from the mental health database. The mapping relationship therein can be one-to-one or one-to-many. For example, the number of floating-point operations, the instruction execution efficiency, the memory bandwidth, and the data throughput respectively form a mapping set with the weight factors corresponding to the number of floating-point operations, the weight factors corresponding to the instruction execution efficiency, the weight factors corresponding to the memory bandwidth, and the weight factors corresponding to the data throughput preset in the mental health database. Substituting the real-time number of floating-point operations, the instruction execution efficiency, the memory bandwidth, and the data throughput into the mapping set, the weight factors corresponding to the number of floating-point operations, the weight factors corresponding to the instruction execution efficiency, the weight factors corresponding to the memory bandwidth, and the weight factors corresponding to the data throughput are obtained.

[0056] In this embodiment, through the multivariate analysis of the number of floating-point operations, instruction execution efficiency, video memory bandwidth, and data throughput, the correlation between these parameters is considered. The number of floating-point operations is positively correlated with the video memory bandwidth. When the number of floating-point operations is high, the video memory needs to be accessed frequently. If the video memory bandwidth is insufficient, it will cause the computing unit to be idle, forming a "memory wall" bottleneck, which has a negative impact on the initial index of the execution quality of the GPU. When the instruction execution efficiency is high but the video memory bandwidth is low and the data supply is not timely, the instructions cannot be executed due to waiting for data, reducing the execution efficiency of the GPU. On the contrary, when the video memory bandwidth is high and the instruction execution efficiency is low, although the data transmission speed is fast but the processing speed cannot keep up, the performance of the GPU cannot be fully utilized, affecting the execution quality. The data throughput is jointly affected by the number of floating-point operations, instruction execution efficiency, and video memory bandwidth. When all three of these indicators are at a relatively high level, the GPU can quickly process a large number of tasks, increasing the data throughput and improving the execution quality. If any one of these indicators is low, it will limit the data throughput and lead to a decline in the execution quality.

[0057] Specifically, the compatible complexity of each GPU is obtained through matching. The specific matching process is as follows:

[0058] Match the initial execution quality indicators of each GPU with the compatible complexity corresponding to each predefined initial execution quality indicator interval to determine the intervals to which the initial execution quality indicators of each GPU belong, and obtain the compatible complexity of each interval and assign it to the GPU corresponding to the initial execution quality indicator. Finally, the compatible complexity of each GPU is obtained.

[0059] The psychological data analysis and warning module is used to perform data processing on the comprehensive complexity of each psychological data slice and the compatible complexity of each GPU to complete the allocation of psychological data slices and GPUs. Finally, the mental health monitoring platform analyzes and warns the student psychological data set.

[0060] Furthermore, to complete the allocation of psychological data slices and GPUs, the specific analysis process is as follows:

[0061] Perform a difference process on the comprehensive complexity of each psychological data slice and the compatible complexity of each GPU respectively to obtain the complexity deviation value between each GPU and each psychological data slice. The GPU corresponding to the minimum complexity deviation value with a certain psychological data slice is recorded as the designated GPU of that psychological data slice. Traverse the complexity deviation values between all GPUs and psychological data slices in sequence to obtain the designated GPU of each psychological data slice, and complete the allocation of psychological data slices and GPUs.

[0062] It should be noted that the complexity deviation value between each GPU and each slice of psychological data may be zero, that is, the comprehensive complexity of a certain slice of psychological data is equal to the compatible complexity of a certain GPU. In addition, if there are negative values among the complexity deviation values between each GPU and each slice of psychological data, that is, the compatible complexity of a certain GPU is less than the comprehensive complexity of a certain slice of psychological data, then this GPU will be discarded to ensure that the complexity deviation values are all natural numbers.

[0063] The mapping relationship between the slices of psychological data and the specified GPU is one-to-one, ensuring that slices with high complexity are assigned to high-performance GPUs and slices with low complexity are assigned to conventional GPUs, greatly avoiding resource waste or overload. At the same time, through the mapping strategy of minimizing the complexity deviation value, the processing time of a single slice on the GPU is reduced. Combined with a real-time adjustment mechanism (such as clock frequency adjustment), it is ensured that the overall processing delay is lower than the threshold. In addition, when the real-time execution quality index of a certain GPU is lower than the threshold (such as abnormal core frequency), the system can quickly remap the slices on this GPU to other devices to ensure the continuity of analysis.

[0064] Specifically, the student psychological data set is analyzed and warned. The specific analysis process is as follows:

[0065] Each slice of psychological data is assigned to the corresponding specified GPU for data analysis, and the total processing delay duration of the mental health monitoring platform is collected. The total processing delay duration can be obtained by extracting from the operation log of the mental health monitoring platform and compared with the predefined total processing delay duration boundary to obtain a comparison result. The mental health monitoring platform determines whether to adjust the processing efficiency of the GPU based on the comparison result. Finally, the mental health monitoring platform analyzes the mental state of the students and visualizes the warning of the analysis results. The mental health monitoring platform outputs the warning results to each receiving terminal it belongs to. Specifically, the mental health monitoring platform extracts the text data of the mental state data of each student and performs keyword retrieval on the text data of the mental state data of each student through the keyword retrieval algorithms built into the mental health monitoring platform, such as the BM algorithm and the hash search algorithm. For example, the keyword retrieval can be anxiety, depression, high stress, shame, bullying, etc., to obtain the keyword retrieval count of each student, and verify it with the predefined keyword retrieval boundary count. If the keyword retrieval count of a certain student is greater than or equal to the keyword retrieval boundary count, then the mental state of this student is warned, and the mental health monitoring platform outputs the warning results to the student terminal, parent terminal, and teacher terminal to which this student belongs, as Figure 2 shown, Figure 2 is a schematic diagram of each receiving terminal belonging to the mental health monitoring platform.

[0066] In a specific embodiment, the analysis of students' mental states can be carried out by obtaining students' interaction data through the Xiaozhi Mental Health Big Data Platform, obtaining students' behavior data through the Anzhi Cloud School Big Data Model, and obtaining students' metadata through the basic data of the Anzhi Cloud Platform; processing students' mental data through multi-modal big data collection and processing technologies (such as the Xiaozhi Model), and providing personalized services for students through artificial intelligence digital human technologies (such as Xiaozhi Classmate); evaluating students' mental data through a professional-level mental evaluation system (such as Xiaozhi Evaluation); comprehensively demonstrating through a multi-factor cross model (such as interaction model, consumption model, moral education model, educational administration model, etc.). Generate an exclusive mental health report for the student side, display indicators such as emotional trends and social activity levels, and provide personalized suggestions (such as "The anxiety index has risen this week. It is recommended to use the meditation function"). On the teacher side, view the overall mental state distribution of the class (such as a heat map), warning list, and intervention records through a visual interface, and support exporting the analysis report. Push a summary of the mental state of the child after desensitization (such as "The child's emotions have fluctuated greatly this week") to the parent side, and provide family education guidance in cooperation with artificial intelligence digital human technologies (such as Xiaozhi Classmate).

[0067] Furthermore, the mental health monitoring platform determines whether to adjust the processing efficiency of the GPU based on the comparison result. The specific adjustment process is as follows:

[0068] If the total processing delay duration of the mental health monitoring platform is greater than or equal to the defined total processing delay duration, the comparison result is recorded as the first comparison result. If the total processing delay duration of the mental health monitoring platform is less than the defined total processing delay duration, the comparison result is recorded as the second comparison result.

[0069] If the comparison result shows the first comparison result, the mental health monitoring platform determines that the processing efficiency of the GPU needs to be adjusted. If the comparison result shows the second comparison result, the mental health monitoring platform determines that the processing efficiency of the GPU does not need to be adjusted.

[0070] Specifically, the adjustment of the processing efficiency of the GPU is as follows:

[0071] Perform a difference operation on the total processing delay duration of the mental health monitoring platform and the defined total processing delay duration to obtain a delay duration deviation value, and match to obtain a clock frequency adjustment value for adjusting the processing efficiency of each GPU. At the same time, real-time monitor the real-time execution quality indicators of each GPU, and compare them with the predefined execution quality threshold. If the real-time execution quality indicator of a certain GPU is less than the execution quality threshold, then based on the difference between the real-time execution quality indicator of the certain GPU and the execution quality threshold, match to obtain the core frequency adjustment value of the GPU for reducing the core frequency of the GPU.

[0072] In a specific embodiment, adjusting the processing efficiency of each GPU specifically includes: setting the initial clock frequency of each GPU to 500 MHz and the clock frequency adjustment value to 200 MHz, then adjusting the real-time clock frequency of each GPU to (500 + 200) MHz to increase the processing efficiency of each GPU. At the same time, reducing the core frequency of the GPU specifically includes: setting the core frequency adjustment value of the GPU to 50 MHz, then adjusting the core frequency of the GPU to (500 + 200 - 50) MHz to reduce the core frequency of the GPU.

[0073] In this embodiment, by matching the delay duration deviation value to adjust the clock frequency, increasing the clock frequency can shorten the instruction execution cycle, accelerate the floating-point operation speed, thereby reducing the slice processing time, significantly reducing the processing delay and improving the real-time performance; by reducing the core frequency of the abnormal GPU, the side effects caused by high-frequency operations can be reduced, the heat dissipation pressure can be effectively relieved, and the performance degradation or hardware damage caused by overheating can be avoided. At the same time, for GPUs whose real-time execution quality index is lower than the threshold, reducing the core frequency can reduce the instruction execution error rate (such as floating-point operation errors) and improve the reliability of task processing.

[0074] The above content is only an example and explanation of the structure of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the described specific embodiments or use similar methods to replace them. As long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should fall within the protection scope of the present invention.

Claims

1. The student psychological data analysis and early warning system based on artificial intelligence is characterized by: include: A data set complexity determination module is used to extract multi-source psychological data sets through the mental health monitoring platform and comprehensively determine the comprehensive complexity of the multi-source psychological data sets; The data slice determination module is used to match the number of psychological data slices according to the comprehensive complexity of the multi-source psychological data set, and finally extract the comprehensive complexity of each psychological data slice; The GPU compatible matching module is used to collect the execution data of each GPU of the mental health monitoring platform in real time, determine the initial indicators of the execution quality of each GPU, and match the compatible complexity of each GPU; The psychological data analysis and early warning module is used to process the comprehensive complexity of each psychological data slice and the compatible complexity of each GPU, complete the allocation of psychological data slices and GPUs, and finally the mental health monitoring platform analyzes and warns the student psychological data set.

2. The student psychological data analysis and early warning system based on artificial intelligence according to claim 1 is characterized by: The comprehensive determination of the comprehensive complexity of the multi-source psychological data set specifically includes: The multi-source psychological data set includes each student's psychological data set, and the sub-complexity of each student's psychological data set is determined and averaged to finally obtain the comprehensive complexity of the multi-source psychological data set; Extracting feature sets based on each student psychological data set, specifically including the total data volume of each student psychological data set, the number of data categories of each student psychological data set, the data time span of each student psychological data set, and the information entropy of each student psychological data set; The total data volume of each student psychological data set, the number of data categories of each student psychological data set, the time span of each student psychological data set and the information entropy of each student psychological data set are weighted aggregated and normalized in turn to obtain the sub-complexity of each student psychological data set.

3. The student psychological data analysis and early warning system based on artificial intelligence according to claim 2 is characterized by: The matching obtains the number of psychological data slices, and the specific matching process is: The comprehensive complexity of the multi-source psychological data set is matched with the number of psychological data slices corresponding to each predefined comprehensive complexity interval, the interval to which the comprehensive complexity of the multi-source psychological data set belongs is determined, and the number of psychological data slices corresponding to the interval is obtained and used as the psychological data slices of the psychological data set.

4. The student psychological data analysis and early warning system based on artificial intelligence according to claim 1 is characterized by: The comprehensive complexity of extracting each psychological data slice is specifically extracted as follows: According to the number of psychological data slices, the multi-source psychological data set of each student is randomly divided to obtain each psychological data slice; The sub-complexity of each student psychological data set contained in each psychological data slice is averaged to obtain the comprehensive complexity of each psychological data slice.

5. The student psychological data analysis and early warning system based on artificial intelligence according to claim 1 is characterized by: The specific determination process of determining the initial index of execution quality of each GPU is as follows: The execution data of each GPU specifically includes the number of floating-point operations of each GPU in a quality determination period, the instruction execution efficiency of each GPU in a quality determination period, the video memory bandwidth of each GPU in a quality determination period, and the data throughput of each GPU in a quality determination period; The floating point operation limit number, instruction execution limit efficiency, video memory limit bandwidth and limit data throughput were extracted from the mental health database; The degree of deviation between the number of floating-point operations of each GPU in the quality determination cycle and the defined number of floating-point operations, the degree of deviation between the instruction execution efficiency of each GPU in the quality determination cycle and the defined instruction execution efficiency, the degree of deviation between the memory bandwidth of each GPU in the quality determination cycle and the defined memory bandwidth, and the degree of deviation between the data throughput of each GPU in the quality determination cycle and the defined data throughput are weighted and aggregated in sequence to obtain the initial execution quality index of each GPU.

6. The student psychological data analysis and early warning system based on artificial intelligence according to claim 5 is characterized by: The matching obtains the compatible complexity of each GPU, and the specific matching process is: The initial execution quality index of each GPU is matched with the compatible complexity corresponding to the predefined initial execution quality index intervals, the intervals to which the initial execution quality index of each GPU belongs are determined, the compatible complexity of each interval is obtained and allocated to the GPU corresponding to the initial execution quality index, and finally the compatible complexity of each GPU is obtained.

7. The student psychological data analysis and early warning system based on artificial intelligence according to claim 1 is characterized by: The above-mentioned psychological data slices and GPU allocation are completed. The specific analysis process is as follows: The comprehensive complexity of each psychological data slice and the compatible complexity of each GPU are respectively subjected to difference processing to obtain the complexity deviation value between each GPU and each psychological data slice. The GPU corresponding to the minimum complexity deviation value between a certain psychological data slice is recorded as the designated GPU of the psychological data slice. The complexity deviation values ​​between all GPUs and psychological data slices are traversed in turn to obtain the designated GPU of each psychological data slice, thus completing the allocation of psychological data slices and GPUs.

8. The student psychological data analysis and early warning system based on artificial intelligence according to claim 1 is characterized by: The specific analysis process of analyzing and warning the student psychological data set is as follows: Each psychological data slice is assigned to the corresponding designated GPU for data analysis, and the total processing delay time of the mental health monitoring platform is collected and compared with the predefined total processing delay definition time to obtain a comparison result. The mental health monitoring platform determines whether to adjust the processing efficiency of the GPU based on the comparison result. Finally, the mental health monitoring platform analyzes the students' psychological state and visualizes the analysis results as a warning. The mental health monitoring platform outputs the warning results to its receiving terminals.

9. The student psychological data analysis and early warning system based on artificial intelligence according to claim 8 is characterized by: The mental health monitoring platform determines whether to adjust the processing efficiency of the GPU based on the comparison results. The specific adjustment process is as follows: If the total processing delay duration of the mental health monitoring platform is greater than or equal to the total processing delay definition duration, the comparison result is recorded as the first comparison result; if the total processing delay duration of the mental health monitoring platform is less than the total processing delay definition duration, the comparison result is recorded as the second comparison result; If the comparison result shows the first comparison result, the mental health monitoring platform determines that the processing efficiency of the GPU needs to be adjusted. If the comparison result shows the second comparison result, the mental health monitoring platform determines that the processing efficiency of the GPU does not need to be adjusted.

10. The student psychological data analysis and early warning system based on artificial intelligence according to claim 9 is characterized by: The processing efficiency of the GPU is adjusted, and the specific adjustment process is as follows: The total processing delay time of the mental health monitoring platform and the total processing delay definition time are subtracted to obtain the delay time deviation value, and the clock frequency adjustment value is matched to adjust the processing efficiency of each GPU. At the same time, the execution quality real-time indicator of each GPU is monitored in real time and compared with the predefined execution quality threshold. If the execution quality real-time indicator of a GPU is less than the execution quality threshold, the core frequency adjustment value of the GPU is matched according to the difference between the execution quality real-time indicator of the GPU and the execution quality threshold, which is used to reduce the core frequency of the GPU.

Images