Student psychological state recognition method and system based on multidimensional data analysis

Abstract

The application relates to the technical field of data processing, and particularly discloses a student psychological state recognition method and system based on multidimensional data analysis, which realizes the precision and efficiency of student mental health management. First, data cleaning and standardization processing are integrated to eliminate invalid information and unify data dimensions, thereby significantly improving analysis reliability. Through a collaborative analysis mechanism, different dimensional data features can be dynamically associated, and the adaptive adjustment function of the feature extraction process can automatically optimize algorithm weights according to data quality, thereby ensuring the extraction accuracy of key psychological indicators. Through data-driven closed-loop management, mental health monitoring is changed from passive coping to active prevention, the efficiency of school psychological crisis intervention is effectively improved, scientific support is provided for educational decision-making, and the precise allocation of mental health resources is promoted.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a student psychological state recognition method and system based on multi-dimensional data analysis. BACKGROUND

[0002] With the advancement of education informatization, education informatization proposes to build an "Internet + education" big platform, promote the digitization of the whole process of teaching, management and evaluation, significantly improve the real-time and accuracy of psychological state recognition, and provide technical support for intelligent decision-making in education.

[0003] For example, the invention patent with publication number CN111599472B discloses a method for recognizing the psychological state of students, which comprises: obtaining the feature information of the students to be tested and the relationship between the students to be tested; and recognizing the psychological state of the students to be tested according to the feature information of the students to be tested and the relationship between the students to be tested. The method can recognize the psychological state of the students to be tested according to the feature information of the students to be tested and the relationship between the students to be tested, and improve the accuracy of the psychological state recognition of the students by considering the relationship between the students.

[0004] For example, the invention patent with publication number CN115274105A discloses a processing method, device and equipment for recognizing the state of students, and a storage medium, which comprises: obtaining the card swiping features and learning features of a first student to be tested within a first time; the card swiping features include feature data related to the common card swiping behavior of the campus card of the first student and the campus cards of multiple second students; the learning features include the score rates of multiple subject knowledge points of the first student; inputting the card swiping features of the first student into a pre-trained first prediction model to obtain a first prediction result; inputting the learning features of the first student into a pre-trained second prediction model to obtain a second prediction result; the first prediction result and the second prediction result are respectively used to represent whether the psychological state of the first student is in an abnormal state; if the first prediction result and the second prediction result both represent that the psychological state of the first student is in an abnormal state, a prediction result that the psychological state of the first student is in an abnormal state is output.

[0005] However, in the process of implementing the embodiments of the present application, it is found that the above-mentioned technology at least has the following technical problems: in the existing student psychological state recognition technology, the core bottleneck is the lack of a collaborative fusion mechanism for multi-source heterogeneous data. This data island effect leads to fragmented characteristics in psychological state analysis, on the one hand ignoring the spatio-temporal correlation between different dimensional data, and on the other hand lacking the ability to capture dynamic change patterns, making it difficult to achieve full-link closed-loop optimization from data collection to state recognition, ultimately restricting the accuracy and timeliness of psychological state evaluation. SUMMARY

[0006] In view of the deficiencies of the prior art, the student psychological state recognition method and system based on multi-dimensional data analysis are provided, which can effectively solve the problems involved in the above background art.

[0007] To achieve the above object, the present application is implemented by the following technical solutions: the first aspect of the present application provides a student psychological state recognition method based on multi-dimensional data analysis, comprising: step one, marking a student psychological state management platform as a target platform, collecting student psychological state data of multi-dimensional data sources through the target platform, collecting and analyzing data attribute parameters of each dimensional data source, thereby performing data cleaning on student psychological state data of each dimensional data source; step two, the target platform receives student psychological state data of each dimensional data source after data cleaning, and performs collaborative analysis, obtains and evaluates collaborative analysis process parameters of the target platform, thereby performing data optimization on the collaborative analysis process of the target platform; step three, after data optimization, the target platform extracts student psychological state features from student psychological state data of each dimensional data source, collects and analyzes feature extraction quality indicators, thereby adjusting the student psychological state feature extraction process; step four, the target platform integrates student psychological state features and performs student psychological state feature feedback.

[0008] As a further method, the unqualified data source is cleaned, and the specific cleaning process is: obtaining a data noise abnormality index deviation rate of the unqualified data source and a data type of the unqualified data source; if the data type of the unqualified data source exists in a first type, a data frame insertion density correction value is matched from the multi-dimensional database according to the data noise abnormality index deviation rate of the unqualified data source, thereby performing data cleaning on the first type; if the data type of the unqualified data source exists in a second type, a harmonic number correction value and a reverberation time length correction value are matched from the multi-dimensional database according to the data noise abnormality index deviation rate of the unqualified data source, thereby performing data cleaning on the second type; if the data type of the unqualified data source exists in a third type, a data interpolation density correction value and an abnormal value correction intensity are matched from the multi-dimensional database according to the data noise abnormality index deviation rate of the unqualified data source, thereby performing data cleaning on the third type.

[0009] As a further method, the data of the target platform is optimized, and the specific optimization process is as follows: obtaining the newly added data collaboration task of the target platform, and obtaining the load of each same data processing node to which the target platform belongs, and comparing the load with the defined load; marking the same data processing node with a load less than the defined load as an available data processing node, thereby counting each available data processing node; migrating the data collaboration task of the current data processing node of the target platform to each available data processing node; if there is no available data processing node, setting the newly added data collaboration task of the target platform as the highest priority, and the target platform processes the data corresponding to the highest priority first, thereby completing the data optimization of the data collaboration process of the target platform; obtaining the data collaboration stability factor of the optimized target platform, and performing early warning judgment.

[0010] As a further method, the early warning judgment is performed, and the specific judgment process is as follows: comparing the data collaboration stability factor of the optimized target platform with the data collaboration stability threshold, if the first condition exists, it is determined that no early warning is performed on the data collaboration process of the target platform; if the second condition exists, it is determined that early warning is performed on the data collaboration process of the target platform, and the data noise anomaly threshold correction coefficient is matched from the multidimensional database based on the deviation rate of the data collaboration stability factor of the optimized target platform, thereby correcting the data noise anomaly threshold; the first condition means that the data collaboration stability factor of the optimized target platform is greater than or equal to the data collaboration stability threshold; the second condition means that the data collaboration stability factor of the optimized target platform is less than the data collaboration stability threshold.

[0011] As a further method, it is determined whether to perform feature extraction abnormal feedback, and the specific judgment process is as follows: obtaining the feature recognition accuracy factor growth rate of the target platform, and comparing it with the defined feature recognition accuracy factor growth rate, if the feature recognition accuracy factor growth rate of the target platform is less than the defined feature recognition accuracy factor growth rate, it is determined that the feature extraction abnormal feedback is performed; if the feature recognition accuracy factor growth rate of the target platform is greater than or equal to the defined feature recognition accuracy factor growth rate, the feature extraction iteration number is adjusted upwards to the maximum feature extraction iteration number, and the feature recognition accuracy factor of the target platform after iteration is obtained, if the feature recognition accuracy factor of the target platform after iteration is still less than the feature recognition accuracy threshold, it is determined that the feature extraction abnormal feedback is performed, if the feature recognition accuracy factor of the target platform after iteration is greater than or equal to the feature recognition accuracy threshold, the feature integrity feedback is performed.

[0012] The second aspect of the present application provides a student psychological state recognition system based on multi-dimensional data analysis, comprising: a data cleaning module for marking a student psychological state management platform as a target platform, collecting student psychological state data of multi-dimensional data sources through the target platform, collecting and analyzing data attribute parameters of each dimensional data source, and thus cleaning the student psychological state data of each dimensional data source; a data optimization module for the target platform to receive the student psychological state data of each dimensional data source after data cleaning, and to perform collaborative analysis, obtain and evaluate the collaborative analysis process parameters of the target platform, and thus optimize the collaborative analysis process of the target platform; a process adjustment module for the target platform to extract student psychological state features from the student psychological state data of each dimensional data source after data optimization, collect and analyze feature extraction quality indicators, and thus adjust the student psychological state feature extraction process; and a feature feedback module for the target platform to integrate student psychological state features and perform student psychological state feature feedback.

[0013] Compared with the prior art, the embodiments of the present application have at least the following advantages or beneficial effects:

[0014] (1) The present application provides a student psychological state recognition method and system based on multi-dimensional data analysis, which realizes the precision and efficiency of student psychological health management. First, the invalid information is eliminated and the data dimensions are unified through data cleaning and standardization processing, which significantly improves the analysis reliability. Through the collaborative analysis mechanism, different dimensional data features can be dynamically associated. The adaptive adjustment function of the feature extraction process can automatically optimize the algorithm weight according to the data quality, ensuring the extraction accuracy of key psychological indicators. Through the closed-loop management driven by data, the psychological health monitoring is changed from passive coping to active prevention, effectively improving the efficiency of school psychological crisis intervention, providing scientific support for education decision-making, and promoting the precise allocation of psychological health resources.

[0015] (2) The present application realizes the precision and efficiency of student psychological state analysis through intelligent data cleaning and dynamic regulation mechanism. First, based on the data noise anomaly index, abnormal data sources are automatically selected and classified for cleaning, and interference information is eliminated specifically to ensure data quality. Second, through the delay transmission mechanism, the data type integrity is ensured, the feature recognition accuracy is optimized by matching the correction value. At the same time, the parameters are adjusted differently for different data types (such as the first type of frame insertion correction, the second type of harmonic optimization, and the third type of interpolation adjustment), which significantly improves the cleaning efficiency. Finally, through dynamic parameter adaptive optimization, manual intervention is reduced, and a full-process automated closed-loop management system is constructed to provide high-reliability data support for psychological state recognition.

[0016] (3) The present application significantly improves the operation efficiency and stability of the psychological state analysis system through intelligent load balancing and dynamic task scheduling: firstly, based on real-time comparison of data processing node load, the task is automatically migrated to the idle node, avoiding resource overload and waste, optimizing the allocation of computing resources; secondly, when there is no available resource, the psychological data coordination task is set as the highest priority, ensuring the immediacy of the key analysis process; thirdly, through dynamic comparison of coordination stability factor and threshold, triggering the adaptive warning mechanism, and reversely correcting the noise threshold parameter, forming a closed loop optimization; finally, a flexible resource scheduling system is built, which not only guarantees the data processing efficiency, but also enhances the system's anti-risk ability, realizes the high availability and continuous stability of mental health monitoring service, and provides solid technical support for large-scale psychological state analysis. BRIEF DESCRIPTION OF DRAWINGS

[0017] The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0018] Figure 1 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0019] Figure 2 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0020] Figure 3 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0021] Figure 4 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0022] Figure 5 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0023] Figure 6 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0024] Figure 7 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0025] Figure 8 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0026] Figure 9 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0027] Figure 10 The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. For ordinary skilled in the art, other drawings can be obtained without creative labor on the basis of the following drawings. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, and not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0029] Referring to Figure 1 The first aspect of the present application provides a student psychological state recognition method based on multi-dimensional data analysis, comprising: step one, marking a student psychological state management platform as a target platform, collecting student psychological state data of multi-dimensional data sources through the target platform, collecting and analyzing data attribute parameters of each dimensional data source, so as to clean the student psychological state data of each dimensional data source; step two, the target platform receives the student psychological state data of each dimensional data source after data cleaning, and performs collaborative analysis, obtains and evaluates the collaborative analysis process parameters of the target platform, so as to optimize the data of the collaborative analysis process of the target platform; step three, after data optimization, the target platform extracts student psychological state features from the student psychological state data of each dimensional data source, collects and analyzes feature extraction quality indicators, so as to adjust the student psychological state feature extraction process; step four, the target platform integrates the student psychological state features and performs student psychological state feature feedback.

[0030] In a specific embodiment, the present application realizes the precision and efficiency of student psychological state analysis through intelligent data cleaning and dynamic regulation mechanism: first, based on the data noise anomaly index, the abnormal data sources are automatically screened and classified and cleaned, the interference information is eliminated, and the data quality is guaranteed; second, through the delay transmission mechanism, the data type integrity is ensured, the feature recognition accuracy is optimized by matching the correction value; at the same time, the parameters are adjusted according to the differences of different data types (such as the first type of frame insertion correction, the second type of harmonic optimization, and the third type of interpolation adjustment), which significantly improves the cleaning efficiency; finally, through dynamic parameter self-adaptive optimization, the manual intervention is reduced, and the whole-process automatic closed-loop management system is constructed, which provides high reliability data support for psychological state recognition.

[0031] Specifically, the student psychological state data of each dimensional data source is cleaned, specifically:

[0032] By analyzing the data attribute parameters of each dimensional data source, the data noise anomaly index of each dimensional data source is obtained, and is compared with the data noise anomaly threshold; the above data noise anomaly threshold represents the maximum value allowed by the data noise anomaly index, which is extracted from the multi-dimensional database.

[0033] If the data noise anomaly index of a certain dimension data source is greater than or equal to the data noise anomaly threshold, the dimension data source is marked as an unqualified data source, and data cleaning is performed on the unqualified data source. After cleaning, it is transmitted to the target platform. If the data noise anomaly index of a certain dimension data source is less than the data noise anomaly threshold, the student psychological state data of the dimension data source is marked as qualified data, and is transmitted to the target platform at the transmission time point. The transmission time point refers to the specific time when the target platform uploads the qualified student psychological state data. The time is set by relevant personnel according to actual work needs, target platform operation conditions and other factors. The transmission time point is arranged in turn according to the fixed time interval preset by the relevant personnel.

[0034] At the data transmission time point, the data amount of the student psychological state data in the qualified data is counted and compared with the defined data amount. If the data amount is less than the defined data amount, the data transmission time point is delayed until the data amount is greater than or equal to the defined data amount, and the characteristic recognition precision factor correction value is matched from the multidimensional database according to the total delay time, and the qualified data is transmitted to the target platform. If the data amount is greater than or equal to the defined data amount, the qualified data is directly transmitted to the target platform. The defined data amount represents the minimum value allowed by the data amount, which is extracted from the multidimensional database. The characteristic recognition precision factor correction value represents the proportion value for correcting the characteristic recognition precision factor. The characteristic recognition precision factor correction value is multiplied by the characteristic recognition precision factor, and the result is the characteristic recognition precision factor corrected by the characteristic recognition precision factor correction value. The specific matching process of the characteristic recognition precision factor correction value is as follows: a total delay time-characteristic recognition precision factor correction value mapping table is stored in the multidimensional database. The total delay time is directly queried in the multidimensional database, and the corresponding precision factor characteristic recognition precision factor correction value of the total delay time can be obtained.

[0035] Specifically, the data noise anomaly index of each dimension data source is analyzed as follows: the data attribute parameters of each dimension data source include the data missing value ratio of each dimension data source, the data format error rate of each dimension data source, and the signal-to-noise ratio of the sensor to which each dimension data source belongs; the data missing value ratio refers to the proportion of missing values in the student psychological state data of the data source, reflecting the data integrity. If the student psychological state data is in chart type, the data missing value ratio is specifically the number of missing values ÷ the total number of observed values × 100%. If the student psychological state data is in video type or audio type, the data missing value ratio is specifically the number of missing frames ÷ the total number of observed frames × 100%. The number of missing values or missing frames is counted and the proportion is calculated through a data quality checking tool (such as the pandas library in Python); the data format error rate represents the frequency of format errors in data processing, reflecting the data standardization. The number of format errors ÷ the total number of processing times × 100% can be detected by a format verification program (such as JSON Schema verification) to detect data points or data frames that do not conform to the preset format and calculate the error rate; the signal-to-noise ratio refers to the ratio of signal power to noise power, which measures the quality of the sensor signal. The signal-to-noise ratio is specifically: where Psignal is the signal power and Pnoise is the noise power. The signal and noise power are measured in real time by the sensor data acquisition system, and the signal-to-noise ratio is calculated using a spectrum analysis tool (such as MATLAB).

[0036] The influence of the proportion relationship between the data missing value ratio and the defined data missing value ratio on the data noise anomaly index, the proportion relationship between the data format error rate and the defined data format error rate, and the proportion relationship between the signal-to-noise ratio and the defined signal-to-noise ratio are quantified by the measurement factor, and the influence degrees are combined to obtain the data noise anomaly index.

[0037] A decrease in the sensor signal-to-noise ratio directly leads to an increase in the proportion of noise components in the original signal, making the effective signal overwhelmed, and thus increasing the probability of misjudgment of the data acquisition module for outliers, ultimately resulting in an increase in the data missing value ratio. At the same time, noise interference can damage the integrity of the data structure, such as causing timestamp misalignment and field encoding disorder, thereby indirectly increasing the data format error rate. Conversely, the data format error rate itself can trigger a data recovery mechanism through parsing failure. If the recovery strategy is not perfect (such as simple interpolation), it will further amplify the deviation between the original data and the true value, forming a vicious cycle of "noise-error-missing". As a comprehensive indicator, the data noise anomaly index value will significantly increase with the growth of the missing value ratio, the deterioration of the format error rate, and the attenuation of the signal-to-noise ratio. The three factors together form a driving closed loop of data quality degradation.

[0038] The data noise anomaly index of each dimension data source represents the data noise anomaly degree of each dimension data source, and the specific expression is:

[0039] ;

[0040] In the formula, a is the number of each dimension data source, u is the total number of data source dimensions, is the data noise anomaly index of the a-th dimension data source, is the data missing value ratio of the a-th dimension data source, is the data format error rate of the a-th dimension data source, is the signal-to-noise ratio of the sensor to which the a-th dimension data source belongs, is the preset data missing value ratio in the multidimensional database, is the preset data format error rate in the multidimensional database, is the preset signal-to-noise ratio in the multidimensional database, is the preset data missing value ratio measurement factor in the multidimensional database, is the preset data format error rate measurement factor in the multidimensional database, is the preset signal-to-noise ratio measurement factor in the multidimensional database.

[0041] The above-mentioned data missing value ratio represents the maximum value allowed by the data missing value ratio; the above-mentioned data format error rate represents the maximum value allowed by the data format error rate; and the above-mentioned signal-to-noise ratio represents the minimum value allowed by the signal-to-noise ratio.

[0042] The above-mentioned data missing value ratio measurement factor is used to quantify the influence degree of a unit value of the data missing value ratio on the data noise anomaly index; the above-mentioned data format error rate measurement factor is used to quantify the influence degree of a unit value of the data format error rate on the data noise anomaly index; and the above-mentioned signal-to-noise ratio measurement factor is used to quantify the influence degree of a unit value of the signal-to-noise ratio on the data noise anomaly index. The mapping relationship between the data missing value ratio, the data format error rate, and the signal-to-noise ratio and the corresponding measurement factors in the multidimensional database is that, for example, when the data missing value ratio, the data format error rate, and the signal-to-noise ratio are input into the multidimensional database, the multidimensional database can retrieve the data missing value ratio measurement factor, the data format error rate measurement factor, and the signal-to-noise ratio measurement factor, and the value range of each is between 0 and 1.

[0043] Furthermore, data cleaning is performed on the unqualified data source. The specific cleaning process is: obtaining the data noise anomaly index deviation rate of the unqualified data source and the data type of the unqualified data source; the data noise anomaly index deviation rate of the unqualified data source refers to obtaining and using the data noise anomaly index of the unqualified data source, subtracting the data noise anomaly threshold, and dividing the result by the data noise anomaly threshold. The final result is the data noise anomaly index deviation rate.

[0044] If the data type of the unqualified data source has a first type, then a data interpolation density correction value is matched from the multidimensional database according to the data noise anomaly index deviation rate of the unqualified data source, thereby performing data cleaning on the first type; the above-mentioned first type represents a video type; the above-mentioned data interpolation density correction value is a proportional coefficient for adjusting the video data interpolation density, and the data interpolation density correction value is multiplied by the current video data interpolation density, and the result is the corrected video data interpolation density, and the corrected video data interpolation density is used to perform data cleaning on the first type; the above-mentioned data interpolation density correction value, the specific matching process is: the data noise anomaly index deviation rate-data interpolation density correction value mapping table is stored in the multidimensional database, and the data noise anomaly index deviation rate is directly queried in the multidimensional database to obtain the data interpolation density correction value corresponding to the data noise anomaly index deviation rate.

[0045] If the data type of the unqualified data source has a second type, the harmonic number correction value and the reverberation time correction value are matched from the multidimensional database according to the data noise anomaly index deviation rate of the unqualified data source, so as to perform data cleaning on the second type; the above-mentioned second type represents the audio type; the above-mentioned harmonic number correction value refers to the proportional value for correcting the number of harmonics in the audio signal, and the harmonic number correction value is multiplied by the number of harmonics in the current audio signal, and the result is the number of harmonics in the corrected audio signal; the above-mentioned reverberation time correction value refers to the proportional value for correcting the reverberation time in the audio signal, and the reverberation time correction value is multiplied by the current The reverberation time in the audio signal is obtained by directly querying the data noise anomaly index deviation rate in the multidimensional database to obtain the harmonic number correction value and the reverberation time correction value corresponding to the data noise anomaly index deviation rate. The result is the reverberation time in the corrected audio signal; the second type of data is cleaned up using the number of harmonics in the corrected audio signal and the reverberation time correction value; the specific matching process of the harmonic number correction value and the reverberation time correction value is as follows: a data noise anomaly index deviation rate-harmonic number correction value mapping table and a data noise anomaly index deviation rate-reverberation time correction value mapping table are stored in the multidimensional database, and the data noise anomaly index deviation rate is directly queried in the multidimensional database to obtain the harmonic number correction value and the reverberation time correction value corresponding to the data noise anomaly index deviation rate.

[0046] If the data type of the unqualified data source exists a third type, a data interpolation density correction value and an abnormal value correction intensity are matched from the multidimensional database according to the data noise anomaly index deviation rate of the unqualified data source, so as to perform data cleaning on the third type; the third type represents a chart type; the data interpolation density correction value represents a proportional value for correcting the data interpolation density, and the data interpolation density correction value is multiplied by the current data interpolation density, and the obtained result is the corrected data interpolation density, wherein the abnormal value correction intensity includes: low intensity: only repairing isolated abnormal values; medium intensity: repairing continuous abnormal segments + trend line assistance, maintaining data coherence; high intensity: full data correction (correction window = full segment) and three sigma principle detection, ensuring data rationality; the third type is cleaned by using the corrected data interpolation density and the abnormal value correction intensity; the data interpolation density correction value and the abnormal value correction intensity are matched as follows: a data noise anomaly index deviation rate-data interpolation density correction value mapping table and a data noise anomaly index deviation rate-abnormal value correction intensity mapping table are stored in the multidimensional database, and the data noise anomaly index deviation rate is directly queried in the multidimensional database, so that the data interpolation density correction value and the abnormal value correction intensity corresponding to the data noise anomaly index deviation rate are obtained.

[0047] Step two, the target platform receives the student psychological state data of each dimension data source after data cleaning, and performs collaborative analysis to obtain and evaluate the collaborative analysis process parameters of the target platform, so as to optimize the collaborative analysis process of the target platform.

[0048] In one specific embodiment, the present application significantly improves the running efficiency and stability of the psychological state analysis system through intelligent load balancing and dynamic task scheduling: first, based on real-time comparison of data processing node load, tasks are automatically migrated to idle nodes to avoid resource overload and waste and optimize the allocation of computing resources; second, when there is no available resource, the psychological data collaboration task is set as the highest priority to ensure the immediacy of the key analysis process; third, through dynamic comparison of the collaborative stability factor and the threshold value, the adaptive early warning mechanism is triggered, and the noise threshold parameter is corrected in reverse, forming a closed loop optimization; finally, a flexible resource scheduling system is built, which not only guarantees the data processing efficiency, but also enhances the system's anti-risk ability, realizes the high availability and continuous stability of the mental health monitoring service, and provides solid technical support for large-scale psychological state analysis.

[0049] Specifically, the data of the target platform is optimized in the collaborative analysis process, specifically: by evaluating the collaborative analysis process parameters of the target platform, the data collaborative stability factor of the target platform is obtained, and compared with the data collaborative stability factor of the target platform, if the data collaborative stability factor of the target platform is greater than or equal to the data collaborative stability threshold, the student psychological state feature is extracted from the student psychological state data of each dimension data source of the target platform; the above-mentioned data collaborative stability threshold represents the minimum value allowed by the data collaborative stability factor, which is extracted from the multidimensional database.

[0050] If the data collaborative stability factor of the target platform is less than the data collaborative stability threshold, the data of the target platform is optimized in the data collaborative process; the data collaborative stability factor of the target platform represents the stability degree of the data collaborative process of the target platform, and the specific analysis process is: the collaborative analysis process parameters of the target platform include the response abnormal rate of the target platform and the cross-data center synchronization delay time length of the target platform; the above-mentioned response abnormal rate represents the frequency that the target platform fails to respond to the request according to the expected time or mode, which is obtained by calculating the ratio of the number of failed transactions to the total number of transactions, and the number of failed transactions and the total number of transactions can be obtained from the data log of the target platform; the above-mentioned cross-data center synchronization delay time length represents the time length required for data synchronization between different data centers, which is calculated by recording the time stamp of data sending and receiving; for batch synchronization, the average value of multiple operations can be taken, and the key event time stamp (such as transaction start, synchronization completion time) is recorded in the data synchronization process, and the cross-data center synchronization delay time length is calculated through the log.

[0051] The average data noise abnormality index of the data source is obtained, that is, the data noise abnormality indexes of each dimension data source are averaged to obtain the average data noise abnormality index of the data source.

[0052] The influence degree of the proportion relationship between the response abnormal rate and the defined response abnormal rate on the data collaborative stability factor, the influence degree of the proportion relationship between the cross-data center synchronization delay time length and the defined cross-data center synchronization delay time length on the data collaborative stability factor, and the influence degree of the average data noise abnormality index on the data collaborative stability factor are quantified in turn by the measurement factor, and the influence degrees are converged to obtain the data collaborative stability factor of the target platform.

[0053] A high response anomaly rate directly reflects the bottleneck of the system's processing capabilities. When the anomaly rate exceeds the threshold, the task retry mechanism will be triggered, resulting in a backlog of cross-data center synchronization requests in the transmission queue, significantly increasing the synchronization delay. The extension of the synchronization delay will cause data packets to face more complex network fluctuations during transmission, such as packet loss and retransmission or timing disorders caused by link congestion, which will directly push up the average data noise anomaly index. At the same time, the increase in the noise anomaly index will further interfere with the accuracy of the data analysis module. For example, noise data may be misjudged as valid instructions and trigger false responses. Ultimately, these three parameters form a compound weakening effect on the data collaboration stability factor through the triple paths of resource competition, transmission degradation, and semantic interference, resulting in an overall decline in the timeliness, accuracy, and fault tolerance of the platform's collaborative analysis.

[0054] Among them, the data collaborative stability factor of the target platform is specifically expressed as:

[0055] ;

[0056] Where, is the data collaborative stability factor of the target platform, is the response exception rate of the target platform, is the cross-data center synchronization delay of the target platform, is the average data noise anomaly index of the data source, It is the defined response exception rate preset in the multidimensional database. It is a preset definition of cross-data center synchronization delay in the multidimensional database. is the response anomaly rate measurement factor preset in the multidimensional database, It is a preset cross-data center synchronization delay factor in the multidimensional database. It is the preset average data noise anomaly index measurement factor in the multidimensional database.

[0057] The above definition of response exception rate indicates the maximum value allowed for the response exception rate; the above definition of cross-data center synchronization delay duration indicates the maximum value allowed for the cross-data center synchronization delay duration.

[0058] The response abnormality rate metric factor is used to quantify the influence degree of a response abnormality rate unit value on the data collaboration stability factor; the cross-data center synchronization delay time length metric factor is used to quantify the influence degree of a cross-data center synchronization delay time length unit value on the data collaboration stability factor; the average data noise abnormality index metric factor is used to quantify the influence degree of an average data noise abnormality index unit value on the data collaboration stability factor; and the mapping relationship between the response abnormality rate, the cross-data center synchronization delay time length, the average data noise abnormality index, and the corresponding metric factors in the multidimensional database is stored, for example, the response abnormality rate, the cross-data center synchronization delay time length, and the average data noise abnormality index are input into the multidimensional database, and the multidimensional database can retrieve the response abnormality rate metric factor, the cross-data center synchronization delay time length metric factor, and the average data noise abnormality index metric factor, and the value range is between 0 and 1.

[0059] Specifically, the data optimization of the data collaboration process of the target platform is performed, and the specific optimization process is as follows: obtaining a new data collaboration task of the target platform, and simultaneously obtaining the load of each same data processing node to which the target platform belongs, and comparing the load with the defined load; the load refers to the task queue length of the data processing node; the defined load represents the maximum value allowed by the load, which is extracted from the multidimensional database; the load of each same data processing node to which the target platform belongs is obtained from the running log of the target platform; the new data collaboration task refers to a multi-source data fusion processing task triggered by the target platform after receiving the student psychological state data that passes the quality check at a preset time sequence node, which aims to ensure the consistency of cross-system data, realizes the spatio-temporal alignment of structured psychological feature data and unstructured behavior log data through a standardized interface, dynamically allocates computing resources based on the data collaboration stability factor, and finally generates a fusion data set that meets the input specification of the psychological state analysis model, the generation mechanism of such a task is associated with the data transmission time sequence, the quality check result, and the platform load state to form a closed loop, and provides real-time data support for the mental health monitoring system.

[0060] The same data processing node with a load less than the defined load is marked as an available data processing node, and thus each available data processing node is counted; the data collaboration task of the current data processing node of the target platform is migrated to each available data processing node; if there is no available data processing node, the new data collaboration task of the target platform is set as the highest priority, and the target platform processes the data corresponding to the highest priority first, thereby completing the data optimization of the data collaboration process of the target platform; the data collaboration stability factor of the optimized target platform is obtained, and a warning judgment is performed.

[0061] Further, a pre-warning determination is made, and the specific determination process is: comparing the data collaboration stability factor of the optimized target platform with the data collaboration stability threshold, if the first condition exists, determining not to pre-warn the data collaboration process of the target platform, if the second condition exists, determining to pre-warn the data collaboration process of the target platform, and simultaneously obtaining and based on the data collaboration stability factor deviation rate of the optimized target platform, matching the data noise anomaly threshold correction coefficient from the multi-dimensional database, so as to correct the data noise anomaly threshold; the above-mentioned data collaboration stability factor deviation rate refers to the deviation degree between the data collaboration stability factor and the data collaboration stability threshold, specifically, the data collaboration stability threshold is subtracted from the data collaboration stability factor, and the result is divided by the data collaboration stability threshold, and finally the data collaboration stability factor deviation rate is obtained; the above-mentioned data noise anomaly threshold correction coefficient represents the proportion value of correcting the data noise anomaly threshold, the data noise anomaly threshold correction coefficient is multiplied by the data noise anomaly threshold, and the result is the corrected data noise anomaly threshold, wherein the data noise anomaly threshold correction coefficient, the specific matching process is: the multi-dimensional database stores the data collaboration stability factor deviation rate-data noise anomaly threshold correction coefficient mapping table, the data collaboration stability factor deviation rate is directly queried in the multi-dimensional database, and the data collaboration stability factor deviation rate corresponding to the data noise anomaly threshold correction coefficient is obtained; the above-mentioned pre-warning of the data collaboration process of the target platform is specifically notifying the platform manager of the abnormality of the data collaboration process of the target platform through email or short message and the like, which needs to be handled immediately.

[0062] It needs to be explained that when it is determined that the second condition (i.e., the data collaboration stability factor of the optimized target platform is less than the data collaboration stability threshold) is met, it indicates that the current data collaboration process has a stability defect, and the data cleaning quality has not reached the business continuity requirement, so the threshold correction mechanism needs to be started, and the mechanism constructs a negative feedback loop of data quality guarantee through the causal chain of "stability gap-deviation quantization-parameter matching-threshold tightening", when the data collaboration stability factor does not reach the threshold value, the system automatically infers that there is noise interference that has not been completely removed in the data link, so the noise tolerance is improved (the threshold is tightened) to forcibly improve the data cleaning strength.

[0063] The first condition refers to the data collaboration stability factor of the optimized target platform being greater than or equal to the data collaboration stability threshold; the second condition refers to the data collaboration stability factor of the optimized target platform being less than the data collaboration stability threshold.

[0064] Step three, after data optimization, the target platform extracts student psychological state features from student psychological state data in each dimension data source, collects and analyzes feature extraction quality indicators, so as to adjust the student psychological state feature extraction process.

[0065] Specifically, the student psychological state feature extraction process is adjusted, specifically: analyzing the feature extraction quality index to obtain a feature recognition accuracy factor of the target platform, and comparing the feature recognition accuracy factor with a feature recognition accuracy threshold value, if the feature recognition accuracy factor of the target platform is greater than or equal to the feature recognition accuracy threshold value, a feature integrity feedback is performed; the feature recognition accuracy threshold value is used to represent the minimum value allowed by the feature recognition accuracy factor, which is extracted from a multi-dimensional database; the feature integrity feedback refers to generating a feature integrity extraction log and storing it in the target platform.

[0066] If the feature recognition accuracy factor of the target platform is less than the feature recognition accuracy threshold value, the student psychological state feature extraction process is adjusted, and the specific adjustment process is: according to the feature recognition accuracy factor of the target platform, a feature extraction batch correction value and a network channel correction value are matched from the multi-dimensional database, while the feature extraction iteration number is continuously increased, and it is determined whether to perform feature extraction abnormal feedback; the batch correction value refers to a proportional value for correcting the batch size of the data processed by the target platform, and the batch correction value is multiplied by the existing batch size of the processed data, and the result is the batch size of the processed data of the target platform after correction; the network channel correction value refers to a proportional value for correcting the number of network channels of the target platform, and the network channel correction value is multiplied by the existing number of network channels of the target platform, and the result is the number of network channels of the target platform after correction; the specific matching process of the batch correction value and the network channel correction value is: the multi-dimensional database stores a feature recognition accuracy factor-feature extraction batch correction value mapping table and a feature recognition accuracy factor-network channel correction value mapping table, and the feature recognition accuracy factor is directly queried in the multi-dimensional database to obtain the batch correction value and the network channel correction value corresponding to the feature recognition accuracy factor.

[0067] The feature recognition accuracy factor of the target platform represents the accuracy of the target platform in recognizing and extracting student psychological state features, and the specific analysis process is: the feature extraction quality index includes the feature classification accuracy of the target platform, the feature coverage rate of the target platform, and the feature noise ratio of the target platform; the feature classification accuracy measures the correctness of the target platform to the feature classification result, that is, the proportion of correctly classified feature samples to the total feature samples, and the labeled data set is used to output the feature category through the classification model (such as SVM, random forest), and the accuracy rate is calculated by comparing with the real label; the feature coverage rate represents the coverage breadth of the extracted feature set to the problem domain, that is, the proportion of the extracted features to the ideal feature set, which reflects the completeness of the feature engineering, wherein the ideal feature set is listed by relevant personnel, and the extracted features are obtained from the data log of the target platform; the feature noise ratio refers to the ratio of the number of invalid features in the extracted feature set of the target platform to the total number of extracted features, which can be collected from the data log of the target platform.

[0068] According to the average data noise anomaly index of the data source, a first correction value is matched from the multidimensional database; the first correction value refers to a proportional value of the feature recognition precision factor corrected according to the average data noise anomaly index, and the specific matching process is as follows: an average data noise anomaly index-first correction value mapping table is stored in the power database, and the average data noise anomaly index is directly queried in the power database, so that the first correction value corresponding to the average data noise anomaly index is obtained.

[0069] According to the data coordination stability factor of the target platform, a second correction value is matched from the multidimensional database; the second correction value refers to a proportional value of the feature recognition precision factor corrected according to the data coordination stability factor, and the specific matching process is as follows: a data coordination stability factor-second correction value mapping table is stored in the power database, and the data coordination stability factor is directly queried in the power database, so that the second correction value corresponding to the data coordination stability factor is obtained.

[0070] By constructing a double-factor correction mechanism of the data source noise anomaly index and the target platform coordination stability factor, dynamic differentiated regulation of the feature recognition precision factor is realized. Specifically, the first correction value is quantified based on the data noise distribution characteristics, and the second correction value is calculated by synchronously combining the platform multi-source data fusion stability, forming a double-dimensional correction parameter set, generating a feature recognition precision factor adapted to the actual working condition, effectively compensating for the inherent deviation of the data acquisition link and the platform processing architecture, and ensuring that the mapping relationship between the feature recognition precision factor and the real psychological state characteristics remains dynamically consistent.

[0071] By measuring the influence degree of the proportional relationship between the factor quantified feature classification accuracy and the defined feature classification accuracy, the influence degree of the proportional relationship between the feature coverage rate and the defined feature coverage rate, and the influence degree of the proportional relationship between the feature noise ratio and the defined feature noise ratio on the feature recognition precision factor, the influence degrees are coupled, and the coupling results are corrected by using the first correction value and the second correction value, so as to obtain the feature recognition precision factor of the target platform.

[0072] The reduction of feature noise ratio (i.e. noise level reduction) can directly improve the feature classification accuracy, as the effective signal highlights make the classifier more easily distinguish the boundary and reduce misjudgment; at the same time, the low noise environment suppresses the generation of redundant features, so that the system can capture high value features preferentially on the premise of maintaining high feature coverage, avoiding the invalid occupation of computing resources by "low-quality coverage", and the feature coverage will decrease due to the missed detection of real features, which indirectly reduces the identification ability of the classifier to edge cases. Finally, through the closed-loop effect of "noise interference classification → classification affecting coverage → coverage feeding back to noise", the three parameters form a complex constraint on the feature recognition precision factor, and the imbalance of any link will lead to the difficulty of achieving the optimal balance between the comprehensiveness, purity and distinguishability of feature recognition.

[0073] wherein the feature recognition precision factor of the target platform is specifically expressed as:

[0074] ;

[0075] wherein, the feature recognition precision factor of the target platform, the first correction value, the second correction value, the feature classification accuracy of the target platform, the feature coverage of the target platform, the feature noise ratio of the target platform, the preset defined feature classification accuracy in the multidimensional database, the preset defined feature coverage in the multidimensional database, the preset defined feature noise ratio in the multidimensional database, the preset feature classification accuracy measurement factor in the multidimensional database, the preset feature coverage measurement factor in the multidimensional database, the preset feature noise ratio measurement factor in the multidimensional database.

[0076] The above defined feature classification accuracy represents the minimum value allowed by the feature classification accuracy; the above defined feature coverage represents the minimum value allowed by the feature coverage; and the above defined feature noise ratio represents the maximum value allowed by the feature noise ratio.

[0077] The feature classification accuracy rate measurement factor is used to quantify the influence degree of the feature classification accuracy rate unit value on the feature recognition accuracy factor; the feature coverage rate measurement factor is used to quantify the influence degree of the feature coverage rate unit value on the feature recognition accuracy factor; the feature noise ratio measurement factor is used to quantify the influence degree of the feature noise ratio unit value on the feature recognition accuracy factor; and the mapping relationship between the feature classification accuracy rate, the feature coverage rate, the feature noise ratio and the corresponding measurement factors in the multidimensional database is stored, for example, the feature classification accuracy rate, the feature coverage rate and the feature noise ratio are input into the multidimensional database, and the multidimensional database can retrieve the feature classification accuracy rate measurement factor, the feature coverage rate measurement factor and the feature noise ratio measurement factor, and the value range is between 0 and 1.

[0078] Further, it is determined whether to perform feature extraction abnormal feedback, and the specific determination process is: obtaining the feature recognition accuracy factor growth rate of the target platform, and comparing it with the defined feature recognition accuracy factor growth rate, if the feature recognition accuracy factor growth rate of the target platform is less than the defined feature recognition accuracy factor growth rate, it is determined to perform feature extraction abnormal feedback; the defined feature recognition accuracy factor growth rate represents the minimum value allowed by the feature recognition accuracy factor growth rate, which is extracted from the multidimensional database; the feature recognition accuracy factor growth rate of the target platform refers to the growth value of the feature recognition accuracy factor of the target platform per unit time, which can be extracted from the data log of the target platform.

[0079] If the feature recognition accuracy factor growth rate of the target platform is greater than or equal to the defined feature recognition accuracy factor growth rate, the feature extraction iteration number is adjusted upward to the maximum feature extraction iteration number, and the feature recognition accuracy factor of the target platform after iteration is obtained, if the feature recognition accuracy factor of the target platform after iteration is still less than the feature recognition accuracy threshold, it is determined to perform feature extraction abnormal feedback, if the feature recognition accuracy factor of the target platform after iteration is greater than or equal to the feature recognition accuracy threshold, feature integrity feedback is performed.

[0080] The feature extraction abnormal feedback refers to reminding the platform manager that the feature recognition accuracy degree of the target platform cannot meet the actual use through short message or email and the like.

[0081] Step four, the target platform integrates the student psychological state feature, and performs student psychological state feature feedback; the specific feedback refers to visualizing the student psychological state feature.

[0082] In one specific embodiment, the present application provides a student psychological state recognition method and system based on multi-dimensional data analysis, which realizes the precision and efficiency of student mental health management. First, the data is cleaned and standardized, which can eliminate invalid information and unify the data dimension, significantly improving the analysis reliability. Through the collaborative analysis mechanism, different dimensional data features can be dynamically associated. The adaptive adjustment function of the feature extraction process can automatically optimize the algorithm weight according to the data quality, ensuring the extraction accuracy of key psychological indicators. Through data-driven closed-loop management, the psychological health monitoring is changed from passive response to active prevention, effectively improving the efficiency of school psychological crisis intervention, providing scientific support for education decision-making, and promoting the precise allocation of mental health resources.

[0083] Referring to Figure 2 The second aspect of the present application provides a student psychological state recognition system based on multi-dimensional data analysis, which includes a data cleaning module, a data optimization module, a process adjustment module, a feature feedback module and a multi-dimensional database.

[0084] The multi-dimensional database is used to store the parameters involved in the student psychological state recognition system based on multi-dimensional data analysis.

[0085] The data cleaning module is connected to the data optimization module and the process adjustment module, the data optimization module is connected to the process adjustment module, the process adjustment module is connected to the feature feedback module, and the data cleaning module, the data optimization module and the process adjustment module are all connected to the multi-dimensional database.

[0086] The data cleaning module is used to mark the student psychological state management platform as a target platform, collect student psychological state data of multi-dimensional data sources through the target platform, and analyze the data attribute parameters of each dimensional data source, so as to clean the student psychological state data of each dimensional data source.

[0087] The data optimization module is used to receive the student psychological state data of each dimensional data source after data cleaning by the target platform, and perform collaborative analysis to obtain and evaluate the collaborative analysis process parameters of the target platform, so as to optimize the collaborative analysis process of the target platform.

[0088] The process adjustment module is used to extract student psychological state features from the student psychological state data of each dimensional data source after data optimization by the target platform, collect and analyze the feature extraction quality indicators, and adjust the student psychological state feature extraction process.

[0089] The feature feedback module is used to integrate the student psychological state features by the target platform, and perform student psychological state feature feedback.

[0090] Figure 3For the data acquisition and cleaning process chart of the application, in the data acquisition and cleaning process, first, the student psychological state data of the multi-dimensional data source is collected, and the data attribute parameters of each dimensional data source are analyzed, then, the data noise abnormality index is obtained according to the analysis result, and it is compared with the preset data noise abnormality threshold, if the data noise abnormality index is greater than or equal to the threshold, the dimensional data source is marked as unqualified data source; if it is less than the threshold, it is marked as qualified data, for the unqualified data source, further obtain its data noise abnormality index deviation rate and data type, and according to the data type, match the corresponding cleaning correction value for data cleaning, after cleaning, transmit to the target platform, at the same time, the data amount of qualified data is counted, and compared with the defined data amount, if the data amount is less than the defined data amount, the data transmission time point is delayed, the total delay time is obtained, the feature recognition precision factor correction value is matched from the multi-dimensional database, finally, the data acquisition and cleaning process is ended, and the collaborative analysis and data optimization process is prepared.

[0091] Figure 4 For the collaborative analysis and data optimization process chart of the application, the collaborative analysis and data optimization process starts from evaluating the collaborative analysis process parameters of the target platform, focuses on the calculation of the data collaborative stability factor, and compares it with the preset data collaborative stability threshold, if the data collaborative stability factor is greater than or equal to the threshold, it indicates that the collaborative analysis process of the target platform is stable, the student psychological state feature can be directly extracted from the student psychological state data of each dimensional data source, if the data collaborative stability factor is less than the threshold, it indicates that there is a problem in the collaborative analysis process of the target platform, the available data processing nodes need to be identified, and the data collaborative task of the current data processing node is migrated to these nodes, if there is no available node, set the new task as the highest priority and process it first, after completing the data collaborative optimization, the data collaborative stability factor of the target platform is obtained again, and the early warning judgment is carried out, finally, the collaborative analysis and data optimization process is ended, and the feature extraction process is prepared.

[0092] Figure 5For the feature extraction flowchart of the present application, the feature extraction flowchart starts from calculating the feature recognition accuracy factor of the target platform, which is an important indicator to measure the quality of feature extraction. The feature recognition accuracy factor is compared with the preset feature recognition accuracy threshold. If it is greater than or equal to the threshold, it indicates that the quality of feature extraction is good, and the feature integrity feedback can be performed. If it is less than the threshold, it indicates that there is a problem with the quality of feature extraction, and the feature extraction batch correction value and network channel correction value need to be matched from the multi-dimensional database according to the feature recognition accuracy factor, and the feature extraction iteration number is continuously increased. In the iteration process, the growth rate of the feature recognition accuracy factor is obtained in real time, and is compared with the defined feature recognition accuracy factor growth rate. If the growth rate is less than the defined growth rate, the feature extraction abnormal feedback is performed. If it is greater than or equal to the defined growth rate, the iteration number is adjusted to the maximum, and the feature recognition accuracy factor after iteration is obtained again for comparison. If it is still less than the threshold, the feature extraction abnormal feedback is continued. If it is greater than or equal to the threshold, it indicates that the quality of feature extraction has met the requirements, and the feature integrity feedback can be performed. Finally, the student psychological state features are integrated, and the student psychological state feature feedback is performed, and the whole process is ended.

[0093] Figure 6 For the home page interface diagram of the student psychological state management platform of the present application, the clear layout of the psychological state management platform is displayed, which provides an entrance for users to quickly understand the core functions and key information of the platform, and facilitates users to quickly jump to different function modules. Figure 7 For the first interface diagram of the abnormal feature of the student psychological state management platform of the present application, Figure 8 For the second interface diagram of the abnormal feature of the student psychological state management platform of the present application, the emotional distribution state of students is displayed. Through carefully designed charts, data visualization elements, etc., the emotional state of students is presented in an intuitive way. For example, different colors represent different emotions, and the proportion of different emotions in students is clearly displayed, so that the abnormal features of students' emotions can be quickly found, such as the proportion of a certain negative emotion being too high, so that targeted intervention measures can be taken in time to protect the mental health of students. Figure 9 For the data detail interface diagram of the student psychological state management platform of the present application, the distribution details of the data accepted by the platform are displayed in detail, so that users can deeply understand the composition and source of the data. Figure 10 For the abnormal feedback interface diagram of the student psychological state management platform of the present application, the abnormal state of students and the platform is displayed, which provides an important basis for timely discovering and solving students' psychological problems and ensuring the stable operation of the platform, so as to ensure that the student psychological state management work can be carried out efficiently and orderly.

[0094] The above is only an example and description of the structure of the present application, and those skilled in the art can make various modifications or supplements to the described specific embodiments or replace them with similar ways, as long as they do not deviate from the structure of the present application or exceed the scope defined by the present application, which shall belong to the protection scope of the present application.

Claims

1. A system for identifying student psychological states based on multidimensional data analysis, characterized by: include: The data cleaning module is used to mark the student psychological state management platform as the target platform, collect student psychological state data from multi-dimensional data sources through the target platform, collect and analyze the data attribute parameters of each dimensional data source, and thus clean the student psychological state data from each dimensional data source; The data optimization module is used for the target platform to receive the student psychological state data from various dimensional data sources after data cleaning, and to conduct collaborative analysis, obtain and evaluate the collaborative analysis process parameters of the target platform, so as to optimize the data of the collaborative analysis process of the target platform; The process adjustment module is used to extract student psychological state features from student psychological state data from various dimensional data sources after data optimization, collect and analyze feature extraction quality indicators, and thus adjust the student psychological state feature extraction process; The feature feedback module is used to integrate students’ psychological state features into the target platform and provide feedback on students’ psychological state features; The student psychological state identification method based on multidimensional data analysis includes: Step 1: Mark the student psychological state management platform as the target platform, collect student psychological state data from multi-dimensional data sources through the target platform, collect and analyze the data attribute parameters of each dimensional data source, and then perform data cleaning on the student psychological state data from each dimensional data source; Step 2: The target platform receives the student psychological state data from various dimensional data sources after data cleaning, performs collaborative analysis, obtains and evaluates the collaborative analysis process parameters of the target platform, and thus optimizes the data of the collaborative analysis process of the target platform; Step 3: After data optimization, the target platform extracts student psychological state features from student psychological state data from various dimensional data sources, collects and analyzes feature extraction quality indicators, and thus adjusts the student psychological state feature extraction process; Step 4: The target platform integrates students’ psychological state characteristics and provides feedback on students’ psychological state characteristics; The data cleaning of student psychological state data from various dimensional data sources specifically refers to: By analyzing the data attribute parameters of each dimensional data source, the data noise anomaly index of each dimensional data source is obtained and compared with the data noise anomaly threshold; If the data noise anomaly index of a certain dimensional data source is greater than or equal to the data noise anomaly threshold, the dimensional data source is marked as an unqualified data source, and the unqualified data source is cleaned and transmitted to the target platform after cleaning. If the data noise anomaly index of a certain dimensional data source is less than the data noise anomaly threshold, the student psychological state data of the dimensional data source is marked as qualified data and transmitted to the target platform at the transmission time; At the data transmission time point, the amount of student psychological state data in the qualified data is counted and compared with the defined data amount. If the data amount is less than the defined data amount, the data transmission time point is delayed until the data amount is greater than or equal to the defined data amount, and the feature recognition precision factor correction value is obtained and matched from the multidimensional database based on the total delay time, and the qualified data is transmitted to the target platform. If the data amount is greater than or equal to the defined data amount, the qualified data is directly transmitted to the target platform; The data noise anomaly index of each dimensional data source represents the degree of data noise anomaly of each dimensional data source. The specific expression is: ； In the formula, a is the number of the data source of each dimension, , u is the total number of data source dimensions, is the data noise anomaly index of the a-th dimension data source, is the ratio of missing values ​​of the data source of the a-th dimension, is the data format error rate of the a-th dimension data source, is the signal-to-noise ratio of the sensor to which the a-th dimension data source belongs, It is the missing value ratio of defined data preset in the multidimensional database. The error rate of defined data formats preset in multidimensional databases, is the preset defined signal-to-noise ratio in the multidimensional database, It is a preset data missing value ratio measurement factor in the multidimensional database. It is the data format error rate measurement factor preset in the multidimensional database. It is a preset signal-to-noise ratio measurement factor in the multidimensional database; The defined data missing value ratio represents the maximum value allowed by the data missing value ratio; the defined data format error rate represents the maximum value allowed by the data format error rate; and the defined signal-to-noise ratio represents the minimum value allowed by the signal-to-noise ratio; The data optimization of the collaborative analysis process of the target platform specifically refers to: By evaluating the collaborative analysis process parameters of the target platform, the data collaborative stability factor of the target platform is obtained and compared with the data collaborative stability threshold of the target platform. If the data collaborative stability factor of the target platform is greater than or equal to the data collaborative stability threshold, the target platform extracts student psychological state characteristics from the student psychological state data of each dimensional data source; If the data collaboration stability factor of the target platform is less than the data collaboration stability threshold, data optimization is performed on the data collaboration process of the target platform; The data collaborative stability factor of the target platform is specifically expressed as: ； Where, is the data collaborative stability factor of the target platform, is the response exception rate of the target platform, is the cross-data center synchronization delay of the target platform, is the average data noise anomaly index of the data source, It is the defined response exception rate preset in the multidimensional database. It is a preset definition of cross-data center synchronization delay in the multidimensional database. is the response anomaly rate measurement factor preset in the multidimensional database, It is a preset cross-data center synchronization delay factor in the multidimensional database. It is the preset average data noise anomaly index measurement factor in the multidimensional database; The definition of the response abnormality rate indicates the maximum value allowed for the response abnormality rate; the definition of the cross-data center synchronization delay duration indicates the maximum value allowed for the cross-data center synchronization delay duration; The aforementioned adjustment of the student psychological state feature extraction process specifically refers to: Analyze the feature extraction quality indicators to obtain the feature recognition accuracy factor of the target platform and compare it with the feature recognition accuracy threshold. If the feature recognition accuracy factor of the target platform is greater than or equal to the feature recognition accuracy threshold, then provide feature integrity feedback. If the feature recognition accuracy factor of the target platform is less than the feature recognition accuracy threshold, the student psychological state feature extraction process is adjusted. The specific adjustment process is as follows: according to the feature recognition accuracy factor of the target platform, the feature extraction batch correction value and the network channel correction value are matched from the multidimensional database, and the number of feature extraction iterations is continuously increased. It is also determined whether to provide feature extraction abnormality feedback. The feature recognition accuracy factor of the target platform is specifically expressed as: ； Where, Identify precise factors for the characteristics of the target platform, is the first correction value, is the second correction value, is the feature classification accuracy of the target platform, is the feature coverage of the target platform, is the characteristic noise ratio of the target platform, is the classification accuracy of the predefined features in the multidimensional database, is the coverage of defined features preset in the multidimensional database, is the defined characteristic noise ratio preset in the multidimensional database, It is a feature classification accuracy measurement factor preset in the multidimensional database. is a feature coverage measurement factor preset in a multidimensional database. It is a characteristic noise ratio measurement factor preset in a multidimensional database; The defined feature classification accuracy rate represents the minimum value allowed for the feature classification accuracy rate; the defined feature coverage rate represents the minimum value allowed for the feature coverage rate; and the defined feature noise ratio represents the maximum value allowed for the feature noise ratio.

2. The system according to claim 1, wherein: The data cleaning of unqualified data sources is carried out as follows: Obtain the data noise anomaly index deviation rate of the unqualified data source and the data type of the unqualified data source; If the data type of the unqualified data source includes a first type, a data insertion density correction value is matched from the multidimensional database according to the data noise anomaly index deviation rate of the unqualified data source, thereby performing data cleaning on the first type; If the data type of the unqualified data source exists in the second type, then the harmonic quantity correction value and the reverberation time correction value are matched from the multidimensional database according to the data noise anomaly index deviation rate of the unqualified data source, thereby performing data cleaning on the second type; If the data type of the unqualified data source exists in a third type, then the data interpolation density correction value and the outlier correction strength are matched from the multidimensional database according to the data noise anomaly index deviation rate of the unqualified data source, so as to perform data cleaning on the third type; The first type represents a video type; the data interpolation density correction value is a proportional coefficient used to adjust the video data interpolation density. The data interpolation density correction value is multiplied by the current video data interpolation density to obtain the corrected video data interpolation density. The corrected video data interpolation density is used to perform data cleaning on the first type; the data interpolation density correction value is specifically matched by storing a data noise anomaly index deviation rate-data interpolation density correction value mapping table in a multidimensional database, and directly querying the data noise anomaly index deviation rate in the multidimensional database to obtain the data interpolation density correction value corresponding to the data noise anomaly index deviation rate; The second type represents an audio type; the harmonic number correction value refers to a proportional value for correcting the number of harmonics in the audio signal, and the harmonic number correction value is multiplied by the number of harmonics in the current audio signal, and the result is the corrected number of harmonics in the audio signal; the reverberation time correction value refers to a proportional value for correcting the reverberation time in the audio signal, and the reverberation time correction value is multiplied by the reverberation time in the current audio signal, and the result is the corrected reverberation time in the audio signal; data of the second type is cleaned using the corrected number of harmonics in the audio signal and the corrected reverberation time in the audio signal; The harmonic number correction value and the reverberation time correction value are matched in the following manner: a data noise anomaly index deviation rate-harmonic number correction value mapping table and a data noise anomaly index deviation rate-reverberation time correction value mapping table are stored in a multidimensional database, and the data noise anomaly index deviation rate is directly queried in the multidimensional database to obtain the harmonic number correction value and the reverberation time correction value corresponding to the data noise anomaly index deviation rate; The third type represents a chart type; the data interpolation density correction value represents a proportional value for correcting the data interpolation density. The data interpolation density correction value is multiplied by the current data interpolation density, and the result is the corrected data interpolation density. The outlier correction strength includes: low strength: only repairing isolated outliers; medium strength: repairing continuous outlier segments + trend line assistance to maintain data consistency; high strength: full data correction and three sigma principle detection to ensure data rationality; the third type is cleaned up using the corrected data interpolation density and outlier correction strength; the data interpolation density correction value and the outlier correction strength, the specific matching process is: storing a data noise anomaly index deviation rate-data interpolation density correction value mapping table and a data noise anomaly index deviation rate-outlier correction strength mapping table in a multidimensional database, and directly querying the data noise anomaly index deviation rate in the multidimensional database to obtain the data interpolation density correction value and outlier correction strength corresponding to the data noise anomaly index deviation rate.

3. The system according to claim 1, wherein: The specific analysis process of the data noise anomaly index of each dimensional data source is as follows: The data attribute parameters of each dimensional data source include the data missing value ratio of each dimensional data source, the data format error rate of each dimensional data source, and the signal-to-noise ratio of the sensor to which each dimensional data source belongs; The influence of the proportional relationship between the data missing value ratio and the defined data missing value ratio on the data noise anomaly index, the influence of the proportional relationship between the data format error rate and the defined data format error rate on the data noise anomaly index, and the influence of the proportional relationship between the signal-to-noise ratio and the defined signal-to-noise ratio on the data noise anomaly index are quantified in turn by measurement factors, and the data noise anomaly index is obtained by combining the various influence degrees. The data noise anomaly index of each dimensional data source represents the degree of data noise anomaly of each dimensional data source.

4. The system according to claim 1, wherein: The data collaboration stability factor of the target platform characterizes the stability of the data collaboration process of the target platform. The specific analysis process is as follows: The collaborative analysis process parameters of the target platform include the response anomaly rate of the target platform and the cross-data center synchronization delay duration of the target platform; The average data noise anomaly index of the data source is obtained. The degree of influence of the proportional relationship between the response anomaly rate and the defined response anomaly rate on the data collaboration stability factor, the degree of influence of the proportional relationship between the cross-data center synchronization delay time and the defined cross-data center synchronization delay time on the data collaboration stability factor, and the degree of influence of the average data noise anomaly index on the data collaboration stability factor are quantified in turn through measurement factors. The various degrees of influence are aggregated to obtain the data collaboration stability factor of the target platform.

5. The system according to claim 4, characterized in that: The data collaboration process of the target platform is optimized, and the specific optimization process is as follows: Obtain the newly added data collaboration tasks of the target platform, and at the same time obtain the load of each data processing node belonging to the target platform, and compare them with the defined load; Mark the data processing nodes of the same category whose load is less than the defined load as available data processing nodes, thereby counting the available data processing nodes; Migrate the data collaboration tasks of the current data processing nodes of the target platform to each available data processing node; If there are no available data processing nodes, the newly added data collaboration task of the target platform is set to the highest priority, and the target platform first processes the data corresponding to the highest priority, thereby completing the data optimization of the data collaboration process of the target platform; Obtain the data collaborative stability factor of the optimized target platform and make early warning judgments.

6. The system according to claim 5, characterized in that: The specific process of making early warning judgment is as follows: Compare the optimized data collaboration stability factor of the target platform with the data collaboration stability threshold. If the first condition exists, it is determined that no warning will be issued for the data collaboration process of the target platform. If the second condition exists, it is determined that an early warning will be issued for the data collaboration process of the target platform. At the same time, the deviation rate of the data collaboration stability factor of the optimized target platform is obtained and matched from the multidimensional database to obtain the data noise anomaly threshold correction coefficient, thereby correcting the data noise anomaly threshold. The first condition refers to that the data collaborative stability factor of the optimized target platform is greater than or equal to the data collaborative stability threshold; The second condition refers to that the data collaborative stability factor of the optimized target platform is less than the data collaborative stability threshold.

7. The system according to claim 1, wherein: The target platform’s feature recognition accuracy factor represents the accuracy of the target platform in identifying and extracting students’ psychological state features. The specific analysis process is as follows: The feature extraction quality indicators include the feature classification accuracy of the target platform, the feature coverage of the target platform, and the feature noise ratio of the target platform; Matching a first correction value from a multidimensional database according to an average data noise anomaly index of a data source; According to the data collaborative stability factor of the target platform, a second correction value is matched from the multidimensional database; By quantifying the degree of influence of the proportional relationship between the feature classification accuracy and the defined feature classification accuracy on the feature recognition precision factor, the degree of influence of the proportional relationship between the feature coverage and the defined feature coverage on the feature recognition precision factor, and the degree of influence of the proportional relationship between the feature noise ratio and the defined feature noise ratio on the feature recognition precision factor through measurement factors, each degree of influence is coupled, and the coupling result is corrected using the first correction value and the second correction value, so as to obtain the feature recognition precision factor of the target platform.

8. The system according to claim 1, wherein: The specific process of determining whether to perform feature extraction abnormality feedback is as follows: Obtain the growth rate of the feature recognition precision factor of the target platform and compare it with the growth rate of the defined feature recognition precision factor. If the growth rate of the feature recognition precision factor of the target platform is less than the growth rate of the defined feature recognition precision factor, it is determined that feature extraction abnormality feedback is performed; If the growth rate of the feature recognition precision factor of the target platform is greater than or equal to the defined feature recognition precision factor growth rate, the number of feature extraction iterations will be adjusted upward to the maximum number of feature extraction iterations, and the feature recognition precision factor of the target platform after the iteration is completed will be obtained. If the feature recognition precision factor of the target platform after the iteration is completed is still less than the feature recognition precision threshold, it will be determined to perform feature extraction abnormality feedback. If the feature recognition precision factor of the target platform after the iteration is completed is greater than or equal to the feature recognition precision threshold, feature integrity feedback will be performed.

Images