Emotion data analysis method based on dynamic iteration
By adopting dynamic iterative methods in sentiment analysis, setting multiple monitoring indicator categories and optimizing sentiment models, the problems of low emotion recognition accuracy and poor adaptability in the existing technology are solved, and more efficient employee sentiment management is achieved.
Patent Information
- Application Number
- CN202510251727.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-04
- Publication Date
- 2025-06-13
AI Technical Summary
The existing single fixed emotion analysis model is prone to misjudgment when multiple people’s emotions are recognized, and cannot adapt to changes in individual emotions at different periods, affecting corporate management efficiency.
Using a sentiment data analysis method based on dynamic iteration, multiple monitoring indicator categories are set and a first-level analysis model is established to improve the emotion recognition accuracy through the fusion analysis of multi-dimensional parameters. Monitoring data packets are collected periodically and the first-level emotion model is optimized to adapt to individual personality differences.
It improves the accuracy of employee emotional recognition, reduces misjudgments caused by personality differences, and enhances corporate management efficiency.
Smart Images
Figure CN120145308A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of emotion analysis, and particularly to a method for emotion data analysis based on dynamic iteration. Background Art
[0002] Emotion is a state that synthesizes a person's feelings, thoughts, and behaviors, and plays an important role in interpersonal communication. At present, more and more enterprises monitor the emotional states of employees and schedule the overall work tasks in a timely manner based on the emotional states of employees, thereby improving the work efficiency of the enterprises.
[0003] However, at present, when analyzing the emotion data of employees, a single fixed emotion analysis model is often used. On the one hand, when a single emotion recognition model is used to recognize the emotions of multiple people, due to the different personality characteristics of different people, it is easy to misjudge the emotional states of each person. On the other hand, since a single person's emotional expression may also vary at different times, a fixed emotion analysis model cannot be adaptively corrected in a timely manner, and it is also easy to misjudge the emotional state of the person. This affects the management efficiency of the enterprise. Summary of the Invention
[0004] The purpose of this application is: to solve the above technical problems, this application provides a method for emotion data analysis based on dynamic iteration, aiming to improve the accuracy of emotion recognition for employees.
[0005] In some embodiments of this application, according to the different ways of expressing emotions, multiple monitoring index categories are set (for example, facial expressions, body movements, speaking tones, body electrical signals, etc.), and a first-level analysis model for each monitoring index category is established, and the accuracy of emotion recognition for the person to be monitored is improved through the fusion analysis of multi-dimensional parameters.
[0006] In some embodiments of this application, by periodically collecting the monitoring data packets of each person to be monitored, the first-level emotion models of each person to be monitored are continuously optimized to improve the adaptability of each first-level emotion model to the corresponding person to be monitored, and to avoid the problem of large errors in the single emotion recognition model when recognizing the emotions of multiple people due to the personality differences of each person to be monitored, so as to achieve the accurate recognition of the emotions of each person to be monitored.
[0007] In some embodiments of this application, a method for emotion data analysis based on dynamic iteration is provided, including: Traverse historical data to establish multiple monitoring index categories and establish a benchmark emotion model; Obtain the training data packets of each person to be monitored, and generate the first-level emotion models of each person to be monitored according to all the training data packets and the benchmark emotion model; Obtain the monitoring data packets of each person to be monitored according to the preset update time node, and generate the iterative strategies of each first-level emotion model based on all the monitoring data packets; Among them, it also includes: Establish a sequence of persons to be monitored A, A = (a 1 , a 2 … a i … a n ), where a i is the i-th person to be monitored; n is the number of persons to be monitored.
[0008] In some embodiments of the present application, when establishing the benchmark emotion model, it includes: Establish a sequence of monitoring index categories B, B = (b 1 , b 2 … b i … b m ), where b i is the i-th type of monitoring index; m is the number of monitoring index categories; Generate multiple training data packets according to historical data; Generate the first-level analysis models of each monitoring index category based on all the training data packets; Establish a sequence of first-level analysis models P, P = (p 1 , p 2 … p i … p m ), where pi is the first-level analysis model of the i-th type of monitoring index; Establish an initial weight sequence R, R = (r 1 , r 2 … r i … r m ), where r i is the initial weight of the i-th first-level analysis model; and ( r i = 1); Establish the benchmark emotion model according to the initial weight sequence R and the sequence of first-level analysis models P.
[0009] In some embodiments of the present application, when generating the first-level analysis models of each monitoring index category, it includes: Set the i-th type of monitoring index as the target monitoring index in turn according to the monitoring index sequence B; Set multiple characteristic parameters of the target monitoring index and establish a preprocessing model of the target monitoring index; Generate a characteristic parameter - emotion evaluation value mapping table of the target monitoring index based on all the training data packets; Generate the first-level analysis model of the target monitoring index according to the characteristic parameter - emotion evaluation value mapping table and the preprocessing model; Generate the first-level analysis models for each monitoring index in sequence.
[0010] In some embodiments of the present application, generating the first-level emotion models for each person to be monitored includes: Set a i as the target person to be monitored according to the sequence of numbers A of the persons to be monitored; Obtain the training data packet of the target person to be monitored; Generate the first-level weight coefficient sequences R1, R1 = (r 11 , r 12 … r 1i … r 1m ) for the target person to be monitored according to the patrol data packet, where r 1i is the first-level weight coefficient of the i-th first-level analysis model; and ( r 1i = 1); Generate the first comparison model according to the first-level weight coefficient sequence; Generate the corrected evaluation value f according to the first-level weight coefficient sequence R1 and the initial weight coefficient R; Preset the first corrected evaluation value threshold F1; If f < F1, set the reference emotion model as the first-level emotion model of the target person to be monitored; If f > F1, generate the corrected weight coefficient sequence R2, R2 = (r 21 , r 22 … r 2i … r 2m ); and ( r 2i = 1); r 2i = (r i + r 1i ) / 2; Generate the first-level emotion model of the target person to be monitored according to the corrected weight coefficient sequence and the first-level analysis model sequence P; Generate the first-level emotion models of each person to be monitored in sequence; Establish the first-level emotion model sequence J, J = (j 1 , j 2 … j i … j n ), where j i is the first-level emotion model of the i-th person to be monitored.
[0011] In some embodiments of the present application, generating the corrected evaluation value f includes: f = e1 * Q1 * (r i - r 1i )2] + e2 * Q2 * [(U1 - U2)2 : Among them, e1 is a preset first weight coefficient; e2 is a preset second weight coefficient; Q1 is a preset first fixed coefficient; Q2 is a preset second fixed coefficient; U1 is an emotion evaluation value of the target person to be monitored generated according to the benchmark emotion model and the training data packet of the target person to be monitored; U2 is an emotion evaluation value of the target person to be monitored generated according to the first comparison model and the training data packet of the person to be monitored.
[0012] In some embodiments of the present application, the iteration strategy of each first-level emotion model is generated according to all monitoring data packets, including: Obtain the monitoring data packet of the target person to be monitored at the current update time node: Set the first-level emotion model of the target person to be monitored as the target first-level emotion model; Generate multiple groups of validation set data according to the monitoring data packet; Generate a credibility evaluation value g of the target first-level emotion model at the current update time node according to all the validation set data; Generate the credibility evaluation values of the corresponding first-level emotion models of each person to be monitored at the current update time node in sequence; Establish a sequence G of credibility evaluation values at the current update time node, G = (g 1 , g 2 …g i …j n ), where j i is the credibility evaluation value of the first-level emotion model corresponding to the i-th person to be monitored at the current update time node; Set the iteration strategy of each first-level emotion model according to the sequence G of credibility evaluation values.
[0013] In some embodiments of the present application, generating the credibility evaluation value g of the second comparison model at the current update time node includes: g = e3 * Q3 * (d i - d') 2 + e4 * Q4 * d'; e3 is a preset third weight coefficient; e4 is a preset fourth weight coefficient; Q3 is a preset third fixed coefficient; Q4 is a preset fourth fixed coefficient; is the number of groups of validation set data; d i is the confidence value of the second comparison model generated based on the i-th group of validation set data; d' is the average confidence value.
[0014] In some embodiments of the present application, when setting the iteration strategy of each first-level emotion model, it includes: Preset a first credibility evaluation value threshold G1 and a second credibility evaluation value threshold G2, and G1 < G2; If gi < G1, set the iteration strategy of the i-th first-level emotion model at the current update time node as the first-level iteration strategy; If G1 < gi < G2, set the iteration strategy of the i-th first-level emotion model at the current update time node as the second-level iteration strategy; If g > G2, set that the i-th first-level emotion model does not generate an iteration strategy at the current update time node.
[0015] In some embodiments of the present application, the first-level iteration strategy includes: If the iteration strategy of the i-th first-level emotion model at the current update time node is the first-level iteration strategy; Set the i-th first-level emotion model as the first-level model to be corrected; Obtain the monitoring data packet of the person to be monitored corresponding to the first-level model to be corrected at the current update time node; Generate multiple optimized data packets according to the monitoring data packet; Correct each first-level analysis model in the first-level model to be corrected according to all the optimized data packets.
[0016] In some embodiments of the present application, the second-level iteration strategy includes: If the iteration strategy of the i-th first-level emotion model at the current update time node is the first-level iteration strategy; Set the i-th first-level emotion model as the second-level model to be corrected; Obtain the monitoring data packet of the person to be monitored corresponding to the first-level model to be corrected at the current update time node; Generate a second-level weight coefficient sequence R2, R2 = (r 21 , r 22 …r 2i …r 2m ), where r 2i is the second-level weight coefficient of the i-th first-level analysis model generated based on the monitoring data packet; and ( r 2i = 1); Correct the second-level model to be corrected according to the second-level weight coefficient sequence R2.
[0017] Compared with the prior art, the beneficial effect of an emotion data analysis method based on dynamic iteration in an embodiment of the present application is that: In view of different expression manners of emotions, multiple monitoring index categories are set (for example, facial expressions, body movements, speaking tones, body electrical signals, etc.), and first-level analysis models for each monitoring index category are established, and the emotion recognition accuracy of the person to be monitored is improved through the fusion analysis of multi-dimensional parameters.
[0018] By periodically collecting the monitoring data packets of each person to be monitored, continuously optimizing the first-level emotion models of each person to be monitored, improving the adaptability of each first-level emotion model to the corresponding person to be monitored, and avoiding the problem of large errors in the multi-person emotion recognition by a single emotion recognition model due to the personality differences of each person to be monitored, the accurate recognition of the emotions of each person to be monitored is realized. Description of the Drawings
[0019] Figure 1 It is a schematic flowchart of a method for emotion data analysis based on dynamic iteration in a preferred embodiment of the embodiment of the present application. Detailed Embodiments
[0020] The following will further describe in detail the specific embodiments of the present application in conjunction with the drawings and embodiments. The following embodiments are used to illustrate the present application, but are not used to limit the scope of the present application.
[0021] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present application.
[0022] The terms "first" and "second" are only used for descriptive purposes, and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, unless otherwise stated, the meaning of "a plurality" is two or more.
[0023] In the description of the present application, it should be noted that unless otherwise clearly defined and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected, or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific situations.
[0024] As Figure 1 shown, a method for emotion data analysis based on dynamic iteration in a preferred embodiment of the embodiment of the present application includes: S101: Traverse historical data to establish multiple monitoring index categories and establish a benchmark emotion model; S102: Obtain the training data packets of each person to be monitored, and generate the first-level emotion models of each person to be monitored according to all the training data packets and the benchmark emotion model; S103: Obtain the monitoring data packets of each person to be monitored according to the preset update time node, and generate the iterative strategies of each first-level emotion model according to all the monitoring data packets; Among them, it also includes: Establish a sequence A of persons to be monitored, A = (a 1 , a 2 … a i … a n ), where a i is the i-th person to be monitored; n is the number of persons to be monitored.
[0025] Specifically, when establishing the benchmark emotion model, it includes: Establish a sequence B of monitoring index categories, B = (b 1 , b 2 … b i … b m ), where b i is the i-th type of monitoring index; m is the number of monitoring index categories; Generate multiple training data packets according to historical data; Generate the first-level analysis models of each monitoring index category according to all the training data packets; Establish a sequence P of first-level analysis models, P = (p 1 , p 2 … p i … p m ), where pi is the first-level analysis model of the i-th type of monitoring index; Establish an initial weight sequence R, R = (r 1 , r 2 … r i … r m ), where r i is the initial weight of the i-th first-level analysis model; and ( r i = 1); Establish the benchmark emotion model according to the initial weight sequence R and the sequence P of first-level analysis models.
[0026] Specifically, the monitoring index categories include but are not limited to data that can convey human emotions such as facial expressions, body movements, speaking tones, and body electrical signals.
[0027] Specifically, by analyzing historical data, generate the occupancy ratios of each category of monitoring indexes in the historical data, and set the occupancy ratios as the initial weight coefficients of the corresponding first-level analysis models.
[0028] Specifically, by establishing an initial weight coefficient sequence R, the fusion analysis of multi-dimensional data is realized, and the accuracy of emotion diagnosis for the personnel to be monitored is improved.
[0029] It can be understood that in the above embodiments, for different expression ways of emotions, multiple monitoring index categories are set (for example, facial expressions, body movements, speaking tones, body electrical signals, etc.), and a first-level analysis model for each monitoring index category is established. The accuracy of emotion recognition for the personnel to be monitored is improved through the fusion analysis of multi-dimensional parameters.
[0030] Specifically, the time period between two adjacent update time nodes is an iteration period, and the duration of the iteration period is preferably one day.
[0031] Specifically, different acquisition methods are set according to different monitoring index categories. For example, for facial expressions and body movements, video monitoring can be used; for body electrical signals, they can be collected through wearable devices; for sound signals, they can be collected through sound sensors.
[0032] In the preferred embodiment of the present application, when generating the first-level analysis model for each monitoring index category, it includes: Successively set the i-th type of monitoring index as the target monitoring index according to the monitoring index sequence B; Set multiple characteristic parameters of the target monitoring index, and establish a preprocessing model for the target monitoring index; Generate a mapping table of characteristic parameter - emotion evaluation value for the target monitoring index according to all training data packets; Generate the first-level analysis model for the target monitoring index according to the mapping table of characteristic parameter - emotion evaluation value and the preprocessing model; Successively generate the first-level analysis models for each monitoring index.
[0033] Specifically, a single first-level analysis model includes a data processing program for this monitoring index, which can extract each characteristic parameter of the monitoring index, and generate an emotion evaluation value corresponding to the personnel reflected by this monitoring index through the analysis of all characteristic parameters.
[0034] Specifically, the larger the emotion evaluation value, the better the overall state of the personnel to be monitored.
[0035] Specifically, generating the first-level emotion model for each personnel to be monitored includes: Successively set a i as the target personnel to be monitored according to the sequence A of personnel to be monitored; Obtain the training data packet of the target personnel to be monitored; Generate the first-level weight coefficient sequences R1 for the target personnel to be monitored according to the inspection data packet, R1=(r 11 , r12 …r 1i …r 1m )), where r 1i is the first-level weight coefficient of the i-th first-level analysis model; and ( r 1i = 1); Generate the first comparison model according to the first-level weight coefficient sequence; Generate the corrected evaluation value f according to the first-level weight coefficient sequence R1 and the initial weight coefficient R; Preset the first corrected evaluation value threshold F1; If f < F1, set the benchmark emotion model as the first-level emotion model of the target person to be monitored; If f > F1, generate the corrected weight coefficient sequence R2 of the target person to be monitored, R2 = (r 21 , r 22 …r 2i …r 2m ); and ( r 2i = 1); r 2i = (r i + r 1i ) / 2; Generate the first-level emotion model of the target person to be monitored according to the corrected weight coefficient sequence and the first-level analysis model sequence P; Generate the first-level emotion models of each person to be monitored in turn; Establish the first-level emotion model sequence J, J = (j 1 , j 2 …j i …j n ), where j i is the first-level emotion model of the i-th person to be monitored.
[0036] Specifically, intercept the historical monitoring data of the target person to be monitored to generate the training data packet of the target person to be monitored, and establish the first-level weight coefficient sequence by analyzing the proportion of various monitoring indicators in the training data packet.
[0037] Specifically, the larger the first-level weight coefficient, the easier it is for the emotion of the target monitoring person to be reflected through the corresponding monitoring indicators. For example, some people like to reflect emotional changes through body movements, and some people like to reflect emotional changes through the tone and intonation of speech.
[0038] Specifically, by weighting each first-level analysis model, establish the first-level emotion model of the target person to be monitored, and improve the accuracy of emotion diagnosis for the target person to be monitored.
[0039] Specifically, generating the corrected evaluation value f includes: f = e1 * Q1 * (r i - r 1i )2] + e2 * Q2 * [(U1 - U2) 2 : Wherein, e1 is a preset first weight coefficient; e2 is a preset second weight coefficient; Q1 is a preset first fixed coefficient; Q2 is a preset second fixed coefficient; U1 is an emotional evaluation value of the target person to be monitored generated according to the benchmark emotion model and the training data packet of the target person to be monitored; U2 is an emotional evaluation value of the target person to be monitored generated according to the first comparison model and the training data packet of the person to be monitored.
[0040] Specifically, all parameters in the model are normalized by the preset first fixed coefficient and second fixed coefficient, so that each parameter is within the same value range.
[0041] Specifically, the larger the corrected evaluation value is, the greater the difference between the emotional expression mode of the target person to be monitored and the corresponding emotional expression mode in the benchmark emotion model is, and the lower the emotion recognition accuracy of the benchmark emotion model for the target person to be monitored is.
[0042] It can be understood that in the above embodiments, the benchmark emotion model is adaptively corrected according to the training data packets of each person to be monitored, so as to generate a first-level emotion model adapted to each person to be monitored, avoiding the problem of large errors in the multi-person emotion recognition by a single emotion recognition model due to the personality differences of each person to be monitored, and realizing the accurate recognition of the emotions of each person to be monitored.
[0043] In the preferred embodiment of the embodiment of the present application, the iterative strategy for generating each first-level emotion model according to all monitoring data packets includes: Obtain the monitoring data packet of the target person to be monitored at the current update time node: Set the first-level emotion model of the target person to be monitored as the target first-level emotion model; Generate multiple groups of validation set data according to the monitoring data packet; Generate a credibility evaluation value g of the target first-level emotion model at the current update time node according to all the validation set data; Generate the credibility evaluation values of the corresponding first-level emotion models of each person to be monitored at the current update time node in turn; Establish a sequence G of credibility evaluation values at the current update time node, G = (g 1 , g 2 … g i … j n ), wherein, j iis the credibility evaluation value of the first-level emotion model corresponding to the i-th person to be monitored at the current update time node; Set the iteration strategy of each first-level emotion model according to the credibility evaluation value sequence G.
[0044] Specifically, generating the credibility evaluation value g of the second comparison model at the current update time node includes: g = e3 * Q3 * (d i - d') 2 + e4 * Q4 * d'; e3 is the preset third weight coefficient; e4 is the preset fourth weight coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; is the number of groups of verification set data; d i is the confidence value of the second comparison model generated based on the i-th group of verification set data; d' is the average confidence value.
[0045] Specifically, generate multiple groups of verification set data by analyzing the monitoring data.
[0046] Specifically, normalize all parameters in the model through the preset third fixed coefficient and fourth fixed coefficient, so that each parameter is within the same value range.
[0047] Specifically, when setting the iteration strategy of each first-level emotion model, it includes: Preset the first credibility evaluation value threshold G1 and the second credibility evaluation value threshold G2, and G1 < G2; If gi < G1, set the iteration strategy of the i-th first-level emotion model at the current update time node as the first-level iteration strategy; If G1 < gi < G2, set the iteration strategy of the i-th first-level emotion model at the current update time node as the second-level iteration strategy; If g > G2, set that the i-th first-level emotion model does not generate an iteration strategy at the current update time node.
[0048] Specifically, the first-level iteration strategy includes: If the iteration strategy of the i-th first-level emotion model at the current update time node is the first-level iteration strategy; Set the i-th first-level emotion model as the first-level model to be corrected; Obtain the monitoring data packet of the person to be monitored corresponding to the first-level model to be corrected at the current update time node; Generate multiple optimized data packets according to the monitoring data packet; Correct each first-level analysis model in the first-level model to be corrected according to all the optimized data packets.
[0049] Specifically, when in the first-level iteration strategy, it indicates that there are relatively large errors in the current first-level emotion model's diagnosis of the emotions of the person to be monitored, which means that the underlying first-level analysis models cannot accurately analyze the monitoring data of the current person to be monitored. It is necessary to promptly adjust the data processing programs of each first-level analysis model to achieve the adaptability of each first-level analysis model to the person to be monitored and improve the accuracy of emotion recognition for the person to be monitored.
[0050] Specifically, the second-level iteration strategy includes: If the iteration strategy of the i-th first-level emotion model at the current update time node is the first-level iteration strategy; Set the i-th first-level emotion model as the second-level model to be corrected; Obtain the monitoring data packet of the person to be monitored corresponding to the first-level model to be corrected at the current update time node; Generate the second-level weight coefficient sequence R2, R2 = (r 21 , r 22 …r 2i …r 2m ), where r 2i is the second-level weight coefficient of the i-th first-level analysis model generated based on the monitoring data packet; and ( r 2i = 1); Correct the second-level model to be corrected according to the second-level weight coefficient sequence R2.
[0051] Specifically, obtain the original weight coefficients of each first-level analysis model in the second-level model to be corrected at the current update time node, select the corresponding second-level weight coefficients in the second-level weight coefficient sequence, and generate the updated weight coefficients of the corresponding first-level analysis model according to the average value of the original weight coefficients and the second-level weight coefficients.
[0052] Specifically, when in the second-level iteration strategy, it indicates that there are certain errors in the current first-level emotion model's diagnosis of the emotions of the person to be monitored. By promptly correcting the weight coefficients of each first-level analysis model in the first-level emotion model, the corrected first-level emotion model will be more in line with the emotion expression mode of the person to be monitored when performing emotion diagnosis.
[0053] It can be understood that in the above embodiments, by performing periodic iterative optimization on each first-level emotion model, the disturbance of the accuracy of the first-level emotion model caused by the differences in the emotion expression modes of a single person at different times can be avoided, and accurate analysis of the emotional states of each person to be monitored can be achieved.
[0054] According to the first concept of the present application, multiple monitoring index categories are set according to different emotional expression methods (for example, facial expressions, body movements, speaking tones, body electrical signals, etc.), and a first-level analysis model for each monitoring index category is established, and the accuracy of emotion recognition for the person to be monitored is improved through the fusion analysis of multi-dimensional parameters.
[0055] According to the second concept of the present application, by periodically collecting the monitoring data packets of each person to be monitored, the first-level emotion models of each person to be monitored are continuously optimized to improve the adaptability of each first-level emotion model to the corresponding person to be monitored, and to avoid the problem of large errors in the single emotion recognition model when performing multi-person emotion recognition due to the personality differences of each person to be monitored, so as to achieve accurate recognition of the emotions of each person to be monitored.
[0056] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present application, several improvements and replacements can be made, and these improvements and replacements should also be regarded as the protection scope of the present application.
Claims
1. A method for analyzing emotional data based on dynamic iteration, characterized in that: Including: Traverse historical data to establish multiple monitoring index categories and establish a benchmark emotion model; Obtain the training data packets of each person to be monitored, and generate the first-level emotion models of each person to be monitored according to all the training data packets and the benchmark emotion model; Obtain the monitoring data packets of each person to be monitored according to the preset update time node, and generate the iterative strategies of each first-level emotion model according to all the monitoring data packets; Among them, it also includes: Establish a sequence of people to be monitored A, A=(a1,a2…a i …a n ), where a i is the i-th person to be monitored; n is the number of people to be monitored.
2. The method for analyzing emotional data based on dynamic iteration according to claim 1, characterized in that: When establishing the benchmark emotion model, it includes: Establish a monitoring indicator category sequence B, B=(b1,b2…b i …b m ), where b i is the i-th monitoring indicator; m is the number of monitoring indicator categories; Generate multiple training data packets according to historical data; Generate the first-level analysis models of each monitoring index category according to all the training data packets; Establish the first-level analysis model series P, P=(p1,p2…p i …p m ), where pi is the primary analysis model of the i-th monitoring indicator; Establish the initial weight sequence R, R=(r1,r2…r i …r m ), where r i is the initial weight of the i-th first-level analysis model; and ( r i =1); Establish a benchmark emotion model according to the initial weight sequence R and the first-level analysis model sequence P.
3. The method for analyzing emotional data based on dynamic iteration according to claim 2, characterized in that: When generating the first-level analysis models of each monitoring index category, it includes: Set the i-th type of monitoring index as the target monitoring index in turn according to the monitoring index sequence B; Set multiple characteristic parameters of the target monitoring index and establish a preprocessing model of the target monitoring index; Generate a characteristic parameter - emotion evaluation value mapping table of the target monitoring index according to all the training data packets; Generate the first-level analysis model of the target monitoring index according to the characteristic parameter - emotion evaluation value mapping table and the preprocessing model; Generate the first-level analysis models of each monitoring index in turn.
4. The method for analyzing emotional data based on dynamic iteration according to claim 3, characterized in that: Generate the first-level emotion models of each person to be monitored, including: According to the number of people to be monitored A, set a i Targeted persons to be monitored; Obtain the training data packet of the target person to be monitored; Generate the first-level weight coefficient sequence R1 of the target person to be monitored according to the inspection data packet, R1=(r 11 , r 12 …r 1i …r 1m ), where r 1i is the first-level weight coefficient of the i-th first-level analysis model; and ( r 1i =1); Generate the first comparison model according to the first-level weight coefficient sequence; Generate the corrected evaluation value f according to the first-level weight coefficient sequence R1 and the initial weight coefficient R; Preset the first corrected evaluation value threshold F1; If f < F1, set the benchmark emotion model as the first-level emotion model of the target person to be monitored; If f>F1, generate the modified weight coefficient series R2 of the target monitored personnel, R2=(r 21 , r 22 …r 2i …r 2m );and( r 2i =1);r 2i =(r i +r 1i ) / 2; Generate the first-level emotion model of the target person to be monitored according to the corrected weight coefficient sequence and the first-level analysis model sequence P; Generate the first-level emotion models of each person to be monitored in turn; Establish a first-level emotion model series J, J=(j1, j2…j i …j n ), where j i is the first-level emotion model of the i-th person to be monitored.
5. The method for analyzing emotional data based on dynamic iteration according to claim 4, characterized in that: Generate the corrected evaluation value f, including: <h2 style=";text-align:left;direction:ltr">f=e1*Q1*[<h2 style=";text-align:left;direction:ltr"> <h2 style=";text-align:left;direction:ltr"> (r<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> -r<h2 style=";text-align:left;direction:ltr"> 1i <h2 style=";text-align:left;direction:ltr"> )2]+e2*Q2*[(U1-U2)<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> ]: Among them, e1 is the preset first weight coefficient; e2 is the preset second weight coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; U1 is the emotion evaluation value of the target person to be monitored generated according to the benchmark emotion model and the training data packet of the target person to be monitored; U2 is the emotion evaluation value of the target person to be monitored generated according to the first comparison model and the training data packet of the person to be monitored.
6. The method for analyzing emotional data based on dynamic iteration according to claim 5, characterized in that: Generate the iterative strategies of each first-level emotion model according to all the monitoring data packets, including: Obtain the monitoring data packet of the target person to be monitored at the current update time node: Set the first-level emotion model of the target person to be monitored as the target first-level emotion model; Generate multiple groups of validation set data according to the monitoring data packet; Generate the credibility evaluation value g of the target first-level emotion model at the current update time node according to all the validation set data; Generate the credibility evaluation values of the corresponding first-level emotion models of each person to be monitored at the current update time node in turn; Establish the trustworthy evaluation value sequence G of the current update time node, G=(g1, g2…g i …j n ), where j i is the trustworthy evaluation value of the first-level emotion model corresponding to the i-th person to be monitored at the current update time node; Set the iterative strategies of each first-level emotion model according to the credibility evaluation value sequence G.
7. The method for analyzing emotional data based on dynamic iteration according to claim 6, characterized in that: Generate the credibility evaluation value g of the second comparison model at the current update time node, including: <h2 style=";text-align:left;direction:ltr">g=e3*Q3*[<h2 style=";text-align:left;direction:ltr"> <h2 style=";text-align:left;direction:ltr"> (d<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> -d')<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> ]+e4*Q4*d'; e3 is the preset third weight coefficient; e4 is the preset fourth weight coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; is the number of groups of validation set data; d i is the confidence value of the second comparison model generated based on the i-th group of validation set data; d' is the average confidence value.
8. The method for analyzing emotional data based on dynamic iteration according to claim 6, characterized in that: When setting the iterative strategies of each first-level emotion model, it includes: Preset the first credibility evaluation value threshold G1 and the second credibility evaluation value threshold G2, and G1 < G2; If gi < G1, set the iteration strategy of the i-th first-level emotion model at the current update time node as the first-level iteration strategy; If G1 < gi < G2, set the iteration strategy of the i-th first-level emotion model at the current update time node as the second-level iteration strategy; If g > G2, set that the i-th first-level emotion model does not generate an iteration strategy at the current update time node.
9. The method for analyzing emotional data based on dynamic iteration according to claim 8, characterized in that: The first-level iteration strategy includes: If the iteration strategy of the i-th first-level emotion model at the current update time node is the first-level iteration strategy; Set the i-th first-level emotion model as the first-level model to be corrected; Obtain the monitoring data packet of the person to be monitored corresponding to the first-level model to be corrected at the current update time node; Generate multiple optimized data packets according to the monitoring data packet; Correct each first-level analysis model in the first-level model to be corrected according to all the optimized data packets.
10. The method for analyzing emotional data based on dynamic iteration according to claim 8, characterized in that: The second-level iteration strategy includes: If the iteration strategy of the i-th first-level emotion model at the current update time node is the first-level iteration strategy; Set the i-th first-level emotion model as the second-level model to be corrected; Obtain the monitoring data packet of the person to be monitored corresponding to the first-level model to be corrected at the current update time node; Generate the secondary weight coefficient series R2, R2=(r 21 , r 22 …r 2i …r 2m ), where r 2i is the secondary weight coefficient of the i-th primary analysis model generated based on the monitoring data packet; and ( r 2i =1); Correct the second-level model to be corrected according to the second-level weight coefficient sequence R2.