A team performance evaluation method and system based on psychological load key feature screening

Abstract

This invention discloses a team performance evaluation method and system based on the screening of key features of psychological load. The invention constructs a highly robust psychological load prediction model by synchronously acquiring multimodal data, combining spatiotemporal coupled feature extraction and multi-objective optimized feature screening. Furthermore, it introduces graph attention mechanisms and dynamic network topology analysis to map individual loads onto the team collaboration network. Using an integral form of the team performance index function, it comprehensively considers the uniformity of load distribution, information flow efficiency, and task stage weights, achieving a dynamic and accurate evaluation of the overall team performance. This solution not only improves the real-time performance and accuracy of team collaboration status monitoring but also provides a visualized decision-making basis for team management, demonstrating significant industrial application value.

Description

Technical Field

[0001] This invention belongs to the field of human factors engineering and artificial intelligence, and in particular relates to a team effectiveness assessment method and system based on screening key features of psychological load. Background Technology

[0002] In modern complex systems and high-risk work environments, team effectiveness is a key factor in determining the success or failure of a task. Traditional team effectiveness assessments mainly rely on post-event questionnaires, subjective observations, or simple performance indicator summaries, which have obvious drawbacks such as being outdated, highly subjective, and lacking in dimensionality.

[0003] With the development of computational social science and human factors engineering, objective assessment methods based on physiological signals (such as EEG and heart rate variability HRV) have gradually become a research hotspot. However, existing related technologies still face the following core challenges: 1. The challenge of synchronizing multi-source heterogeneous data Team collaboration involves multiple modalities, including voice, motion, and physiological signals. Due to the inconsistency of time references among different sensors, existing technologies often struggle to achieve strict time alignment at the microsecond level, leading to artifacts or information loss during feature extraction and severely impacting evaluation accuracy.

[0004] 2. Insufficient robustness of psychological load characteristics Psychological load typically manifests as non-stationary changes in physiological signals. Traditional entropy-based algorithms (such as sample entropy) are sensitive to noise and easily misinterpret motion artifacts as changes in cognitive load, leading to misselection of features and reducing the model's generalization ability.

[0005] 3. Lack of quantitative capture of the "bottleneck effect" Current assessments often focus on average load levels, neglecting the "weakest link" effect in team collaboration. When a member of the team becomes a bottleneck due to overload, overall efficiency drops sharply, but such localized extreme loads are often not effectively identified and warned of by existing static assessment models.

[0006] 4. Staticization of evaluation dimensions Most existing models use discrete time points for evaluation, which cannot reflect the dynamic trajectory of effectiveness evolution over time. Team effectiveness often exhibits emergent characteristics, meaning that it shows different characteristics at specific task phases (such as crisis management periods) than usual, and static models struggle to capture this time-varying feature.

[0007] 5. The blindness of feature selection When processing multimodal high-dimensional data, directly inputting all features can lead to model overfitting and computational burden. Traditional single-index ranking (such as looking only at the Gini index) is prone to getting stuck in local optima and lacks global consideration of feature redundancy and interpretability consistency, resulting in the selected feature set potentially containing distractors.

[0008] To address the aforementioned issues, this invention proposes a novel solution: enhancing signal robustness by introducing spatiotemporal coupling feature extraction, eliminating redundant interference by utilizing multi-objective optimization of recursive features, modeling interaction relationships using dynamic graph neural networks, and finally employing an integral form of the performance index function for full-cycle evaluation, thereby achieving in-depth analysis of team performance. Summary of the Invention

[0009] The machine learning terms involved in this invention are defined as follows: Gini importance refers to the average reduction in the impurity of the prediction result calculated based on the random forest model; SHAP value is a feature attribution value based on game theory, used to explain the contribution of a single feature to a single prediction result.

[0010] To address the shortcomings of the existing technology, this invention provides a team effectiveness assessment method based on screening key features of psychological load, comprising the following steps: Step 1: During the team's execution of the preset tasks, collect each member's EEG, ECG, and interaction logs simultaneously and align them with a unified timestamp to form a multimodal data stream; Step 2: Divide the data stream into segments according to fixed time windows, and extract a subset of low-dimensional key features that simultaneously reflect individual cognitive-physiological states and inter-member interaction behaviors; Step 3: Construct a team state graph with members as nodes and interactions as dynamic edges. Use graph attention mechanism to aggregate neighbor information window by window and output multi-dimensional psychological load vectors of each member in real time. Step 4: Based on the psychological load vector, calculate the nonlinear entropy value of the team load distribution and the task progress efficiency online, and generate a comprehensive performance index that changes over time by integrating the results.

[0011] In step two, the subset of low-dimensional key features is obtained through multi-objective optimization recursive feature elimination. The optimization simultaneously maximizes the Gini importance of features to psychological load, minimizes the variance of cross-member SHAP value distribution and mutual information redundancy, and removes low-scoring features using Pareto front adaptive threshold.

[0012] The EEG features extracted in step two include Theta, Alpha, and Beta energies and phase synchronization based on complex Moray wavelets, and the ECG features include HRV nonlinear dynamics indices with adaptive weighted multiscale fuzzy entropy to suppress motion artifacts.

[0013] In step three, the dynamic edge weights are formed by weighted fusion of communication frequency and BERT semantic similarity, so that the edge weights reflect the effectiveness and relevance of information transmission.

[0014] In step three, the graph attention mechanism is executed after the temporal convolution. First, the TCN captures the temporal dependencies of individuals, and then the GAT module dynamically learns the edge weights, so that the load prediction can simultaneously perceive the individual evolution and the team interaction structure.

[0015] The multi-dimensional psychological load vector output in step three includes at least three dimensions: cognitive load, emotional load, and operational load, making the team's state more comprehensive.

[0016] In step four, the task progress efficiency is weighted by a segmented step function. The function is assigned different importance coefficients according to the "planning-execution-review" stage, so that the performance of the key stage contributes more to the comprehensive efficiency index.

[0017] In step four, the nonlinear entropy value is calculated using Tsallis entropy, with its non-extensibility parameter q>1 to amplify the sensitivity of high-load members to team effectiveness.

[0018] In the fourth step of the integral evaluation, a structural flexibility index is introduced to quantify the team's ability to adapt to sudden disturbances by monitoring the rate of change in network topology connectivity, and this index is used as a negative adjustment factor for performance evaluation.

[0019] This invention also proposes a team efficacy assessment system based on screening key features of psychological load to implement the above method, comprising: A multimodal synchronous acquisition unit is used for concurrent acquisition of EEG, ECG and interaction logs and alignment with a unified timestamp; The feature engineering server deploys a multi-objective optimization recursive feature elimination module and outputs a subset of key features. Load forecasting and network modeling server, deploying GAT-STGCN model, generating psychological load vectors in real time and constructing dynamic team state diagrams; The performance evaluation and visualization server has a built-in nonlinear entropy-efficiency integral function, outputs a comprehensive performance index, and supports backtracking and review.

[0020] Compared with the prior art, the present invention has the following advantages: 1. Multimodal deep fusion and robustness enhancement The spatiotemporal coupled feature extraction module employed in this invention not only extracts the time-frequency domain features of physiological signals but also addresses the instability in feature calculation caused by the non-stationarity of physiological signals by introducing an adaptive weighting function based on the local signal variation coefficient. Compared to traditional methods, this feature extraction logic exhibits stronger noise resistance in complex environments (such as strenuous exercise accompanied by high cognitive load), ensuring the quality of subsequent model input data.

[0021] 2. Intelligent Feature Selection Based on Game Theory The multi-objective optimization recursive feature elimination model described in this invention overcomes the limitations of traditional single-threshold screening. This model uses the NSGA-II algorithm to find the Pareto optimal front, forcibly eliminating features that, while performing well on a single metric, have deficiencies in global redundancy or interpretability. This results in a selected feature subset that not only has high predictive power but also possesses extremely high physical meaning and statistical stability.

[0022] 3. Dynamic network topology and bottleneck early warning In steps S6 and S7, this invention innovatively quantifies psychological load as graph node attributes and dynamically updates edge weights by combining communication frequency and semantic similarity. This dynamic network modeling method can reflect the efficiency of information flow within a team in real time. In particular, by combining the maximum flow algorithm to calculate the bottleneck load index, the system can provide early warning of specific "bottleneck members" before team efficiency declines, a function that existing static evaluation systems cannot achieve.

[0023] 4. Integral-based performance evaluation model The team performance evaluation function of this invention integrates performance over the entire task cycle into a continuous exponent by introducing time integration. Compared to simple weighted summation, the integral form can amplify the impact of long-lasting and far-reaching periods of inefficiency. At the same time, through the design of the exponential function and denominator, it non-linearly penalizes extreme load imbalances (such as one person being overworked while everyone else is idle), thus better conforming to the actual physical laws of human teamwork.

[0024] In summary, this invention constructs a complete closed loop from data acquisition and feature engineering to model prediction and performance evaluation through a series of closely related technical features. It effectively solves the technical bottlenecks in existing team performance evaluation, such as low accuracy, strong lag, and lack of bottleneck early warning, and has extremely high creativity and practicality. Attached Figure Description

[0025] The above and other objects, features, and advantages of exemplary embodiments of the present disclosure will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the present disclosure are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein: Figure 1 This is a flowchart illustrating a team effectiveness assessment method based on screening key features of psychological load according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating a team effectiveness assessment system based on screening key features of psychological load according to an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0027] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0028] It should be understood that although the terms first, second, third, etc., may be used to describe... in the embodiments of the present invention, these... should not be limited to these terms. These terms are only used to distinguish... For example, first... may also be referred to as second... without departing from the scope of the embodiments of the present invention, and similarly, second... may also be referred to as first...

[0029] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0030] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0031] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0032] The optional embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0033] Example 1 like Figure 1 As shown, this invention discloses a team effectiveness assessment method based on screening key features of psychological load, comprising the following steps: Step 1: During the team's execution of the preset tasks, collect each member's EEG, ECG, and interaction logs simultaneously and align them with a unified timestamp to form a multimodal data stream; Step 2: Divide the data stream into segments according to fixed time windows, and extract a subset of low-dimensional key features that simultaneously reflect individual cognitive-physiological states and inter-member interaction behaviors; Step 3: Construct a team state graph with members as nodes and interactions as dynamic edges. Use graph attention mechanism to aggregate neighbor information window by window and output multi-dimensional psychological load vectors of each member in real time. Step 4: Based on the psychological load vector, calculate the nonlinear entropy value of the team load distribution and the task progress efficiency online, and generate a comprehensive performance index that changes over time by integrating the results.

[0034] In step two, the subset of low-dimensional key features is obtained through multi-objective optimization recursive feature elimination. The optimization simultaneously maximizes the Gini importance of features to psychological load, minimizes the variance of cross-member SHAP value distribution and mutual information redundancy, and removes low-scoring features using Pareto front adaptive threshold.

[0035] The EEG features extracted in step two include Theta, Alpha, and Beta energies and phase synchronization based on complex Moray wavelets, and the ECG features include HRV nonlinear dynamics indices with adaptive weighted multiscale fuzzy entropy to suppress motion artifacts.

[0036] In step three, the dynamic edge weights are formed by weighted fusion of communication frequency and BERT semantic similarity, so that the edge weights reflect the effectiveness and relevance of information transmission.

[0037] In step three, the graph attention mechanism is executed after the temporal convolution. First, the TCN captures the temporal dependencies of individuals, and then the GAT module dynamically learns the edge weights, so that the load prediction can simultaneously perceive the individual evolution and the team interaction structure.

[0038] The multi-dimensional psychological load vector output in step three includes at least three dimensions: cognitive load, emotional load, and operational load, making the team's state more comprehensive.

[0039] In step four, the task progress efficiency is weighted by a segmented step function. The function is assigned different importance coefficients according to the "planning-execution-review" stage, so that the performance of the key stage contributes more to the comprehensive efficiency index.

[0040] The segmented step function, based on the task management model, divides the task into planning, execution, and review phases, and assigns importance coefficients to each, such as 1.2, 1.5, and 1.0. These coefficients can be determined through historical data statistics or domain expert experience.

[0041] In step four, the nonlinear entropy value is calculated using Tsallis entropy, with its non-extensibility parameter q>1 to amplify the sensitivity of high-load members to team effectiveness.

[0042] The non-extensibility parameter q ranges from 1.2 to 5.0, preferably from 2.0 to 3.0. By adjusting the q value to be greater than 1, the sensitivity of the entropy function to extreme values ​​in the probability distribution can be enhanced, thereby effectively highlighting the influence of members with high psychological load in the team load distribution.

[0043] In the fourth step of the integral evaluation, a structural flexibility index is introduced to quantify the team's ability to adapt to sudden disturbances by monitoring the rate of change in network topology connectivity, and this index is used as a negative adjustment factor for performance evaluation.

[0044] like Figure 2 As shown, this invention also proposes a team effectiveness assessment system based on screening key features of psychological load to implement the above method, comprising: A multimodal synchronous acquisition unit is used for concurrent acquisition of EEG, ECG and interaction logs and alignment with a unified timestamp; The feature engineering server deploys a multi-objective optimization recursive feature elimination module and outputs a subset of key features. Load forecasting and network modeling server, deploying GAT-STGCN model, generating psychological load vectors in real time and constructing dynamic team state diagrams; The performance evaluation and visualization server has a built-in nonlinear entropy-efficiency integral function, outputs a comprehensive performance index, and supports backtracking and review.

[0045] Example 2 This invention proposes a team effectiveness assessment method based on screening key characteristics of psychological load, comprising the following steps: Step S1: During the team's execution of a pre-set complex collaborative task, a multimodal data synchronous acquisition system is used to concurrently collect the physiological time-series data of each member of the team, including electroencephalogram (EEG) and electrocardiogram (ECG), interactive behavior log data containing operation instructions and communication content, and event tag data that identifies task stages and external environmental interference, through a unified high-precision timestamp, forming the original multi-source dataset. Step S2: Input the physiological time series data into a spatiotemporal coupled feature extraction module, and extract dynamic features representing the cognitive and physiological state of each member to form an individual time series feature sequence; Step S3: Parse the interaction behavior log data and event tag data collected in Step S1, and accurately align them with the individual time-series feature sequence generated in Step S2 using timestamps. Then, segment them along the time axis to generate multimodal data frames containing internal state and external interaction information of individuals in fixed time windows. Aggregate the data frames of all members in all time windows to form an initial high-dimensional feature set at the team level. Step S4: Construct a recursive feature elimination model based on multi-objective optimization and game theory interpretation, and input the initial high-dimensional feature set at the team level into the model; in each iteration, the model comprehensively evaluates the Gini importance of each feature in predicting individual psychological load, the stability of the SHAP value distribution across members, and the mutual information redundancy with other features, and eliminates features whose comprehensive score is lower than the adaptively adjusted Pareto front threshold, until the feature subset converges, and finally outputs an optimal key feature subset of team psychological load; wherein, the iteration termination condition of the recursive feature elimination process is: the size of the feature subset reaches a preset threshold (such as 10%-30% of the initial number of features), or the comprehensive score S of all remaining features reaches a certain threshold. i All are higher than the current adaptive threshold of the Pareto front. Step S5: Input the key feature subset output in step S4 into a pre-trained spatiotemporal graph convolutional network (GAT-STGCN) model based on graph attention mechanism. This model treats team members as graph nodes and the communication and task dependencies between members as dynamically changing graph edges. At each time step, it aggregates the information of neighboring nodes through graph attention mechanism and combines it with its own temporal features to predict the multi-dimensional psychological load vector of each member in the continuous time window. Step S6: Based on the multi-dimensional psychological load vector output in step S5, construct a dynamically weighted team state network, where each network node not only carries the multi-dimensional psychological load vector of an individual, but also has its role identifier in the task attached; the weights of the directed edges in the network are dynamically updated according to the interaction behavior log data parsed in step S3, reflecting the frequency and direction of information flow. Step S7: On the dynamically weighted team state network, a team performance integral evaluation function is applied to calculate the comprehensive performance index (TEI) reflecting the team's overall performance within the task cycle. This function comprehensively quantifies the team's overall load emergence, structural flexibility, and task progress efficiency by integrating a non-linearly coupled expression over time. Its core formula is: The following is an explanation of all characters in the formula: TEI is the final team overall performance index, t0 and t... f Let be the start and end times of the task, respectively, and dt be the time integral infinitesimal. Here, H(P(t)) is a pre-defined positive real-number weighting coefficient used to balance different performance dimensions; H(P(t)) is the generalized entropy of the team load distribution calculated at time point t, used to measure the balance of load distribution; P(t) is the probability distribution vector composed of the psychological load values ​​of all members; and exp(·) is an exponential function. It is the standard deviation of the psychological load values ​​of all members at a given time point. C is the maximum information flow calculated in the dynamic network G(t) considering the psychological load of nodes as a capacity constraint. L (t) is a node capacity vector composed of the reciprocals of the psychological load of each member, and V is the set of nodes in the network. Let d(i,j) be the edge weight from node i to node j, and d(i,j) be the shortest path distance from node i to node j in the network. It is a phase related to the task The relevant piecewise step function is used to adjust the evaluation weights for different task stages. This is the integration operator. As a summation operator, this formula sums the uniformity of the load distribution (numerator H) and the extreme dispersion of the load (denominator e). This method combines the efficiency of network information flow with a term representing the ratio of network information flow efficiency to structural cost, then weights the results according to the task stage, and finally integrates the results over the entire task cycle. This allows for a comprehensive and dynamic evaluation of the team's overall effectiveness. The improvement lies in using time integration instead of discrete summation and introducing nonlinear functions (exponential functions) and metrics deeply coupled with the network structure (maximum flow and path distance). This allows for a more accurate capture of the nonlinear and emergent characteristics of effectiveness evolving over time. Compared to simple linear weighting, this formula can more profoundly reflect the bottleneck effect and resource allocation efficiency in team collaboration.

[0046] The spatiotemporal coupling feature extraction module in step S2 specifically executes the following logic: First, it applies time-frequency analysis based on complex Molay wavelet transform to the input EEG signal to extract energy and phase synchronization features in multiple key frequency bands such as Theta, Alpha, and Beta; second, it locates the R-wave peak of the input ECG signal through continuous wavelet transform and calculates various time-domain, frequency-domain, and nonlinear dynamic indices of heart rate variability (HRV); wherein the nonlinear dynamic indices include a modified, more robust multi-scale fuzzy entropy (MFE) to noise. When calculating the similarity between vectors, this fuzzy entropy introduces an adaptive weighting function based on the local signal variation coefficient to suppress the interference of non-stationary noise such as motion artifacts on the entropy value calculation, thereby more stably quantifying the complexity of the autonomic nervous system.

[0047] Specifically, the recursive feature elimination model based on multi-objective optimization and game theory interpretation in step S4 is implemented as follows: In each iteration, for each feature f to be evaluated... i Calculate a comprehensive score S i Its calculation formula is ,in It is feature f i Average Gini importance in random forest models, CoV(SHAP(f i The SHAP value is the coefficient of variation of the feature's predictions for all members, used to measure the consistency of its interpretability. It is feature f i With feature subset Other features f j Normalized mutual information, These are hyperparameters that control importance, stability, and redundancy penalties, respectively; the model uses the S-parameters of all features. i The values ​​are mapped to a multidimensional objective space, and the non-dominated sorting genetic algorithm (NSGA-II) is used to find the Pareto optimal front. Features located below the front are eliminated, thereby achieving simultaneous optimization of feature performance and redundancy.

[0048] The spatiotemporal graph convolutional network (GAT-STGCN) model for graph attention mechanism in step S5 includes a temporal convolutional module (TCN) and a graph attention convolutional module (GAT). At each time step, the TCN first captures the temporal dependencies of each member's features to generate intermediate hidden states. Then, these hidden states are fed into the GAT module as input, which calculates the attention coefficients between each pair of connected nodes. To learn the edge weights, the calculation method is as follows: Where W is the shared linear transformation weight matrix, These are the hidden state vectors of nodes i and j. This indicates a splicing operation. It is the weight vector of a single-layer feedforward neural network; after aggregating neighbor information through attention weighting, the node representation is updated, so that the prediction of mental load can simultaneously perceive the temporal evolution of individuals and the instantaneous interaction structure of teams.

[0049] In step S6, the weights of the dynamically weighted team state network edges are... The update method is as follows: ,in It represents the communication frequency from member i to member j within the time window. It is the semantic similarity between the information content sent by members within this time window, calculated using a pre-trained BERT model. The method adjusts the weighting coefficients for communication frequency and the importance of information content; it makes the weights of network edges reflect not only the frequency of interaction, but also the effectiveness and relevance of information transmission.

[0050] In step S7, the generalized entropy H(P(t)) of the team load distribution is calculated using the Tsallis entropy, and its formula is as follows: ,in N represents the proportion of members at the k-th discrete psychological load level at time point t. L It is the discrete rank of the psychological load level, and q is a non-extensive parameter; compared with the traditional Shannon entropy, using Tsallis entropy can better capture atypical load distribution characteristics caused by strong correlation or long-range correlation, especially amplifying the proportion of members in extreme states such as "high load" or "low load".

[0051] In step S7, the task phase Related piecewise step functions The definition method is as follows: First, the entire collaborative task is divided into multiple logical stages such as "planning," "execution," and "review"; then, through expert knowledge base or historical data analysis, each stage is further defined. Assign an importance coefficient The function The value of time t represents the current stage of the task. Corresponding coefficients This allows team performance during critical mission phases to contribute more to the final performance index.

[0052] In step S5, the multidimensional psychological load vector includes at least three dimensions: cognitive load dimension, which is quantified by analyzing the Beta band energy of EEG signals and the amplitude of P300 event-related potentials; emotional load dimension, which is quantified by analyzing the low-frequency / high-frequency power ratio (LF / HF) in heart rate variability and the average level of skin conductance; and operational load dimension, which is quantified by analyzing the operation frequency and instruction complexity in the interaction behavior log. This multidimensional definition makes the characterization of psychological load more comprehensive and accurate.

[0053] Example 3 This invention also proposes a team efficacy assessment system based on screening key features of psychological load to implement the above method, comprising: A multimodal data synchronization acquisition unit is equipped with multiple wearable EEG and ECG sensors that are bound to team members, as well as client software for capturing keyboard, mouse operations and voice communication. It also has a built-in unified timestamp service based on the Network Time Protocol (NTP) to timestamp all acquired data streams. A data preprocessing and feature extraction server is communicatively connected to the acquisition unit, and the spatiotemporal coupling feature extraction module is deployed therein to execute feature extraction logic and complete data alignment and segmentation; A feature engineering server, deployed with the aforementioned recursive feature elimination model based on multi-objective optimization and game theory interpretation, is used to receive preprocessed data and perform feature filtering. A load forecasting and network modeling server, deployed with the GAT-STGCN model and dynamic network building module, is used to receive key features and perform steps S5 and S6. A performance evaluation and visualization server, with the team performance score evaluation function built-in, is used to execute step S7, calculate and display the final comprehensive performance index and its sub-indicators in the form of a multi-dimensional dashboard.

[0054] The load forecasting and network modeling server and the performance evaluation and visualization server are deployed on a cloud computing platform that supports GPU acceleration; the GAT-STGCN model is implemented based on the PyTorch Geometric library; the servers transmit data via low-latency, high-throughput streaming data through the gRPC protocol; the performance evaluation and visualization server also includes a backtracking analysis module, which allows users to select any historical time period to reproduce the evolution of the team's network topology, member load distribution, and performance index at that time, in order to support in-depth attribution analysis and decision review.

[0055] Example 4 This disclosure provides a non-volatile computer storage medium storing computer-executable instructions that can perform the steps described in the above embodiments.

[0056] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0057] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0058] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (AN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0059] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0060] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0061] The preferred embodiments of the present invention have been described above to make the spirit of the present invention clearer and easier to understand, and are not intended to limit the present invention. All modifications, substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope summarized by the appended claims.

Claims

1. A team performance evaluation method based on psychological load key feature screening, characterized in that, Includes the following steps: Step 1: During the team's execution of the preset tasks, collect each member's EEG, ECG, and interaction logs simultaneously and align them with a unified timestamp to form a multimodal data stream; Step 2: Divide the data stream into segments according to fixed time windows, and extract a subset of low-dimensional key features that simultaneously reflect individual cognitive-physiological states and inter-member interaction behaviors; Step 3: Construct a team state graph with members as nodes and interactions as dynamic edges, and use graph attention mechanism to aggregate neighbor information window by window to output multi-dimensional psychological load vectors for each member in real time. Step 4: Based on the psychological load vector, calculate the nonlinear entropy value of the team load distribution and the task progress efficiency online, and generate a comprehensive performance index that changes over time by integrating the results.

2. The method of claim 1, wherein, In step two, the subset of low-dimensional key features is obtained through multi-objective optimization recursive feature elimination. The optimization simultaneously maximizes the Gini importance of features to psychological load, minimizes the variance of cross-member SHAP value distribution and mutual information redundancy, and removes low-scoring features using Pareto front adaptive threshold.

3. The method of claim 2, wherein, The EEG features extracted in step two include Theta, Alpha, and Beta energies and phase synchronization based on complex Moray wavelets, and the ECG features include HRV nonlinear dynamics indices with adaptive weighted multiscale fuzzy entropy to suppress motion artifacts.

4. The method of claim 1, wherein, In step three, the dynamic edge weights are formed by weighted fusion of communication frequency and BERT semantic similarity, so that the edge weights reflect the effectiveness and relevance of information transmission.

5. The method as described in claim 1, characterized in that, In step three, the graph attention mechanism is executed after temporal convolution. First, the TCN captures the temporal dependencies of individuals, and then the GAT module dynamically learns the edge weights, enabling load prediction to simultaneously perceive individual evolution and team interaction structure.

6. The method as described in claim 1, characterized in that, The multi-dimensional psychological load vector output in step three includes at least three dimensions: cognitive load, emotional load, and operational load, making the team's state more comprehensive.

7. The method as described in claim 1, characterized in that, In step four, the task progress efficiency is weighted by a segmented step function. This function assigns different importance coefficients according to the "planning-execution-review" stage, so that the performance of the key stages contributes more to the overall efficiency index.

8. The method as described in claim 1, characterized in that, In step four, the nonlinear entropy value is calculated using Tsallis entropy, whose non-extensibility parameter q>1 is used to amplify the sensitivity of high-load members to team effectiveness.

9. The method as described in claim 1, characterized in that, The fourth step of the integral evaluation introduces a structural flexibility index to quantify the team's ability to adapt to sudden disturbances by monitoring the rate of change in network topology connectivity, and this index is used as a negative adjustment factor for performance evaluation.

10. A team efficacy assessment system based on screening key features of psychological load, implementing the method of any one of claims 1-9, comprising: A multimodal synchronous acquisition unit is used for concurrent acquisition of EEG, ECG and interaction logs and alignment with a unified timestamp; The feature engineering server deploys a multi-objective optimization recursive feature elimination module and outputs a subset of key features. Load forecasting and network modeling server, deploying GAT-STGCN model, generating psychological load vectors in real time and constructing dynamic team state diagrams; The performance evaluation and visualization server has a built-in nonlinear entropy-efficiency integral function, outputs a comprehensive performance index, and supports backtracking and review.

Images