Employee work ability assessment method based on multimodal health data fusion

Abstract

The present invention discloses an employee work ability assessment method based on multimodal health data fusion, which includes obtaining multiple health indicator values ​​of physical examination data of the employee to be assessed, multiple self-assessment scale scores of psychological assessment data, and multiple physiological monitoring indicator values ​​of wearable device data; standardizing the acquired data and converting the standardized data into feature vectors respectively; inputting the feature vectors corresponding to the physical examination data and the psychological assessment data into a trained multi-layer perceptron for feature extraction to obtain a physical examination feature vector and an assessment feature vector; inputting the feature vector of the wearable device data into a trained long short-term memory network for feature extraction to obtain a wear feature vector; inputting the physical examination feature vector, the assessment feature vector, and the wear feature vector into a fusion module of a trained multimodal learning model to perform feature fusion, and inputting the fused features into a DNN model of the multimodal learning model to generate a work ability score.

Description

Technical Field

[0001] The present invention relates to enterprise employee work ability assessment technology, and in particular to an employee work ability assessment method based on multimodal health data fusion. Background Art

[0002] With the advancement of the national "Healthy China" strategy, the medical and health service system has gradually shifted from a "treatment-centered" approach to a "health-centered" approach, placing greater emphasis on a model that integrates disease prevention with health promotion. Against this backdrop, companies are increasingly demanding employee health management. Healthy employees not only improve productivity but also reduce medical expenses and human resource management costs.

[0003] In complex work environments, employees' health is crucial to their performance. For example, shale gas exploration encompasses multiple aspects of the business, from equipment operation and drilling operations to field exploration. Employees in different positions not only perform different tasks but also face unique occupational health risks. For example, outdoor explorers spend the majority of their time conducting geological exploration in the field, facing the challenges of extreme weather, rugged terrain, and high physical exertion. This significantly increases their physical health risks, such as burnout, joint injuries, and skin problems. Furthermore, long-term work experiences far from the city and family can lead to social isolation and increased psychological stress, which in turn impacts their mental health and work performance. Therefore, companies need to comprehensively integrate and analyze health information to accurately assess employees' performance. This allows for timely adjustments to positions when performance declines, ensuring the physical and mental health of employees, reducing medical expenses, and ensuring efficient operations to reduce human resource management costs.

[0004] Currently, traditional health management models typically rely on regular physical examinations and health record updates to identify potential health issues. However, these methods often struggle to dynamically assess employees' work capabilities in real time and are subject to lags. Furthermore, existing health management systems typically focus on employees' physical health, neglecting the impact of mental health and work environment on employee performance, making them unable to address diverse work environment demands. Summary of the Invention

[0005] In response to the above-mentioned deficiencies in the prior art, the employee work ability assessment method based on multimodal health data fusion provided by the present invention solves the problem that the prior art cannot accurately assess the work ability of employees in different working environments.

[0006] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0007] A method for evaluating employee work ability based on multimodal health data fusion is provided, which includes the following steps:

[0008] S1. Obtain multiple health indicator values ​​from the employee's physical examination data, multiple self-assessment scale scores from the psychological assessment data, and multiple physiological monitoring indicator values ​​from the wearable device data;

[0009] S2. Standardize the health index values, self-assessment scale scores, and physiological monitoring index values, and convert the standardized physical examination data, wearable device data, and work performance data into feature vectors respectively;

[0010] S3. Input the feature vectors corresponding to the physical examination data and the psychological assessment data into the trained multi-layer perceptron for feature extraction, thereby obtaining a physical examination feature vector and an assessment feature vector;

[0011] S4. Input the feature vector of the wearable device data into the trained long short-term memory network for feature extraction to obtain a wear feature vector;

[0012] S5. Input the physical examination feature vector, assessment feature vector, and wearing feature vector into the fusion module of the trained multimodal learning model to perform feature fusion, and input the fused features into the DNN model of the multimodal learning model to generate a work ability score.

[0013] Furthermore, the method for the fusion module to perform feature fusion includes:

[0014] S51, perform feature splicing on the physical examination feature vector and the wearing feature vector to obtain the first splicing feature :

[0015]

[0016] Among them, Concat() is a concatenation function used to combine two or more feature vectors into one during machine learning or data processing; is the physical examination feature vector; is the wear feature vector;

[0017] S52, use the fully connected layer to the first splicing feature Perform dimensionality reduction operation to obtain the first dimensionality reduction data of 128 dimensions:

[0018]

[0019] in, is the first dimension reduction data; f() is the activation function Relu; is the weight matrix of the physical examination data features; is the bias vector of the first splicing feature;

[0020] S53, perform feature splicing on the evaluation feature vector and the first dimension reduction data to obtain the global feature :

[0021]

[0022] in, is the evaluation feature vector;

[0023] S54, using the fully connected layer to analyze the global features Perform dimensionality reduction to obtain 64-dimensional fused features :

[0024]

[0025] in, is the feature weight matrix after modality fusion; is the bias vector of the global feature.

[0026] The beneficial effect of the above technical solution is that during the data fusion process, instead of concatenating all modal data at once, this solution adopts a phased fusion strategy based on the correlation between the data. First, highly correlated modalities (physical examination feature vectors and wearable feature vectors) are locally fused, and then these fused features are gradually fused with other modal data (assessment feature vectors). This approach avoids the noise and information loss that may be caused by direct data concatenation, better maintains the independence of each modality, and gradually constructs a more effective global feature representation during the subsequent global fusion process.

[0027] Furthermore, the total loss function of the multimodal learning model is expressed as:

[0028]

[0029] in, is the total loss function; is the weight of the i-th feature vector in step S5; L i is the loss value of the i-th feature vector in step S5; and are regularization coefficients; h i and h j are the i-th and j-th feature vectors in step S5 respectively; n is the total number of feature vectors input into the multimodal learning model for each employee in step S5; T is the transpose.

[0030] The beneficial effects of the above technical solution are as follows: the loss function of this solution can dynamically adjust the weight according to the contribution of each modality to the prediction task, thereby improving the prediction performance of the model. In addition, the regularization terms of the second and third terms on the right side of the loss function can effectively prevent a certain modality from monopolizing too much weight, ensuring that the features of each modality are reasonably utilized in the model. At the same time, the modal diversity constraint (the first term on the right side of the loss function) further encourages the differences between different modal features, reduces feature redundancy, and improves the generalization ability of the model. That is, the loss function constructed by this solution can flexibly cope with complex multimodal data scenarios while ensuring feature balance, significantly improving the model's prediction effect on employee work ability.

[0031] Furthermore, the DNN model includes multiple layers of fully connected layers and output layers, and each layer uses a nonlinear activation function ReLU;

[0032] The first hidden layer contains 256 neurons, and its nonlinear activation function ReLU is:

[0033]

[0034] in, is the weight matrix of the first hidden layer, whose dimension is 256×64; is the bias vector of the first hidden layer, whose dimension is 256; is the output of the first hidden layer, whose dimension is 256;

[0035] The second hidden layer contains 128 neurons, and its nonlinear activation function ReLU is:

[0036]

[0037] in, is the weight matrix of the second hidden layer, whose dimension is 128×256; is the bias vector of the second hidden layer, whose dimension is 128; is the output of the second hidden layer, whose dimension is 128;

[0038] The third hidden layer contains 64 neurons, and its nonlinear activation function ReLU is:

[0039]

[0040] in, is the weight matrix of the third hidden layer, whose dimension is 64×128; is the bias vector of the third hidden layer, whose dimension is 64; is the output of the third hidden layer, whose dimension is 64;

[0041] The expression of the output layer is:

[0042]

[0043] in, is the weight matrix of the output layer, whose dimension is 1×64; is the bias vector of the output layer; Rate work ability.

[0044] The beneficial effect of this technical solution is that multimodal data fusion using this method can construct a comprehensive employee health feature vector. This feature vector not only reflects the employee's current health status but also captures changing trends in their health through dynamic updates, thereby more accurately assessing their work ability. Based on the work ability score, company managers can adjust employee work schedules in real time, preventing production risks caused by changes in health status and optimizing human resource management.

[0045] Furthermore, the method for extracting features from the feature vectors of the physical examination data using a multi-layer perceptron includes:

[0046] S31. Use the first connection layer of the multilayer perceptron to map the feature vector of the physical examination data to a high-dimensional space:

[0047]

[0048] in, It is a high-dimensional feature of physical examination data, with a dimension of 128; is the weight matrix of the first connection layer; is the bias term of the first connection layer; is the activation function, To obtain the maximum value function, is the feature vector of the physical examination data;

[0049] S32, use the second connection layer of the multi-layer perceptron to extract the high-dimensional features of the physical examination data to obtain a 64-dimensional physical examination feature vector :

[0050]

[0051] in, is the weight matrix of the second connection layer; is the bias term of the second connection layer;

[0052] S33, use the output layer of the multi-layer perceptron to output the physical examination feature vector ;

[0053] The method of extracting the physical examination feature vector and the evaluation feature vector by the multi-layer perceptron is the same. The output dimension of the first connection layer of the evaluation feature vector is 256, and the output dimension of the second connection layer is 128.

[0054] The method of extracting features from the feature vectors of wearable device data using a long short-term memory network includes:

[0055] S41. Calculate the output of the input gate of the long short-term memory network at the tth time step :

[0056]

[0057] in, 、 are the weight and bias of the input gate respectively; is the hidden state at the t-1th time step; is the M-dimensional time series feature vector of the wearable device data at the tth time step; M is the length of the time window; is the sigmoid function; is the feature vector of M dimension;

[0058] S42. Calculate the output of the forget gate of the long short-term memory network at the tth time step :

[0059]

[0060] in, and are the weight and bias of the forget gate respectively;

[0061] S43. Update the cell state at the tth time step according to the output of the input gate and the output of the forget gate:

[0062] ,

[0063] in, and are the cell states at the tth and t-1th time steps respectively; is the tth time step The parameters calculated from the input of the t-th time step and the hidden state of the previous time step; and They are The weights and biases of is the hyperbolic tangent function;

[0064] S44. Calculate the output of the output gate of the long short-term memory network at the tth time step :

[0065]

[0066] in, and are the weight and bias of the output gate respectively;

[0067] S45. Calculate the features extracted by the long short-term memory network at the tth time step based on the output of the output gate and the unit state :

[0068]

[0069] S46, repeat steps S41 to S45 until t=M, and output the feature vector extracted at the Mth time step , as the wear feature vector.

[0070] The beneficial effects of the above technical solution are: by extracting features from the time series data of wearable devices through the long short-term memory network (LSTM), it can better capture the changes in employees' health status at different time points, and combine physical examination and psychological assessment data to ensure that the generated work ability score is more comprehensive and accurate.

[0071] Furthermore, the training method of the multimodal learning model includes:

[0072] Using the method of steps S1 to S4, obtaining the physical examination feature vectors, assessment feature vectors, and wearing feature vectors of several employees;

[0073] The method of steps S51 to S54 is used to obtain the fused features of each employee and obtain the work ability score of each employee;

[0074] The fused features of each employee are used as the model input features in the training set, and the work ability score of each employee is used as the model output feature in the training set;

[0075] The training set is input into the DNN model in batches, and the DNN model is trained using the adaptive modal weight optimization algorithm AMWO until the model converges or the number of training times reaches the preset number, obtaining a trained DNN model;

[0076] The fusion module and the trained DNN model are used to form a trained multimodal learning model.

[0077] Furthermore, when training the DNN model, the expression for updating the model parameters of the i-th feature vector is:

[0078] ,

[0079] in, and are the model parameters of the i-th feature vector in step S5 during the m+1-th and m-th training times, respectively; for gradient; is the learning rate of the i-th eigenvector in step S5; Hyperparameters adjusted to control the learning rate; is the initial learning rate; is the weight of the i-th feature vector in step S5; L i is the loss value of the i-th feature vector in step S5.

[0080] The beneficial effect of the above technical solution is that, during the process of fusing multimodal data (this solution includes three modalities: physical examination data, wearable device data, and psychological assessment data), different data sources may have different impacts on employee performance. This solution dynamically adjusts the weight of each modality to ensure that the evaluation results accurately reflect the importance of each data source. In other words, this solution optimizes the contribution of each modality based on the actual performance of the data and health status.

[0081] The updated model parameters intelligently handle the complex relationships between multimodal data, preventing a single data source from excessively influencing the evaluation results, and improving the accuracy and robustness of performance assessments. It also adaptively learns the importance of each modal feature, optimizing the evaluation process and avoiding data bias during evaluation.

[0082] Furthermore, the method for obtaining the work ability score includes:

[0083] Obtain multiple health indicator values ​​from employees' physical examination data, scores from multiple self-assessment scales from psychological assessment data, and scores from multiple scoring standard indicators from work performance data;

[0084] According to the scoring table corresponding to the physical examination data, the score of each health indicator value is obtained;

[0085] Based on the scores of health index values, self-assessment scales, and scoring criteria indicators, a weighted average method is used to obtain health data scores, mental health scores, and work performance scores;

[0086] The work ability score is obtained by taking a weighted average approach based on the health data score, mental health score and work performance score.

[0087] Furthermore, multiple health indicators include BMI, blood pressure and blood sugar; multiple self-assessment scales include anxiety self-assessment scale and depression self-assessment scale; multiple scoring standard indicators include performance indicators, attendance rate, work efficiency, teamwork and communication skills and subjective scores of managers; multiple physiological monitoring indicators include heart rate, sleep and activity level.

[0088] Compared with the prior art, the present invention has the following beneficial effects:

[0089] This solution can fully reflect the health status of employees in complex jobs and working environments by integrating multimodal data (physical examination data in physical examination reports, wearable device data collected by wearable devices, and psychological assessment data). This allows employees' health status to be assessed from multiple dimensions. Combined with the multimodal learning model, it can accurately capture the impact of employee health on work ability, thereby obtaining a relatively accurate work ability score for employees based on multimodal data. Based on this score, corporate managers can accurately determine whether employees are currently competent for their jobs, so as to make timely adjustments to their jobs, ensure the physical and mental health of employees, reduce the company's medical expenses, and ensure efficient operation of the company to reduce human resource management costs.

[0090] Employees in different occupations face different occupational health risks. For example, the health assessment of equipment operators focuses on the strain on the musculoskeletal system, while the assessment of prospectors requires greater attention to the physical and psychological stress of working in extreme environments. This solution leverages multimodal data fusion technology and artificial intelligence algorithms to provide personalized work capacity assessments for employees in different occupations, based on the actual work scenarios of different positions (data collected by wearable devices can reflect the impact of different work scenarios on employees' physiological monitoring data). This accurately reflects each employee's specific health status, helps companies identify high-risk employees, optimize health management strategies, and ensure the physical and mental well-being of their employees.

[0091] This solution uses the most recently collected wearable device data for assessments, and the most recently acquired physical examination report and psychological assessment data. This allows for timely capture of changes in employee health and avoids the health risks associated with delayed assessments. Companies can adjust employee work schedules based on the latest work capacity assessment results, preventing a decline in work capacity due to health changes, thereby effectively reducing the risk of production accidents.

[0092] Through accurate work capacity assessments, this solution can effectively reduce work hours lost due to employee health issues. Furthermore, through personalized health interventions and early warnings, companies can reduce the incidence of major illnesses and health issues, thereby reducing high medical expenses. This improves companies' ability to optimize human resource management, reduces economic losses caused by employee health issues, and enhances employee work enthusiasm and long-term work capacity, ultimately improving overall productivity. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 Flowchart of the employee work capacity assessment method based on multimodal health data fusion. DETAILED DESCRIPTION

[0094] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0095] refer to Figure 1 , Figure 1 FIG shows a flow chart of an employee work ability assessment method based on multimodal health data fusion; FIG. Figure 1 As shown, the method S includes steps S1 to S5.

[0096] In step S1, multiple health indicators are obtained from the employee's physical examination data, multiple self-assessment scale scores from the psychological assessment data, and multiple physiological monitoring indicators from the wearable device data. The multimodality in this solution refers to the three modalities of physical examination data, psychological assessment data, and wearable device data. The multiple health indicators include BMI, blood pressure, and blood sugar; the multiple self-assessment scales include anxiety and depression scales; and the multiple physiological monitoring indicators include heart rate, sleep, and activity level.

[0097] In step S2, the health index values, self-assessment scale scores, and physiological monitoring index values ​​are standardized, and the standardized physical examination data, wearable device data, and work performance data are converted into feature vectors respectively; the health index values ​​and self-assessment scale scores are Z-score standardized, and the physiological parameters measured in the wearable device data (such as heart rate, step count, etc.) are Min-Max standardized.

[0098] Vectorize the data after three modal standardization respectively:

[0099] (1) Physical examination report data is usually structured table data, and the value of each health indicator (such as BMI, blood pressure, blood sugar, etc.) is already in numerical form. Assume that the physical examination report contains health indicators (such as BMI, blood pressure, blood sugar, etc.), and the physical examination data is represented as a vector in the following form:

[0100]

[0101] in, is the feature vector of the physical examination data, for -dimensional feature vector; The 1st, 2nd... health indicator values.

[0102] (2) Psychological assessment data is usually a scale score, which indicates the mental health status of employees. Each assessment scale corresponds to multiple scoring items. Assume that the psychological assessment includes There are several scoring items, such as the Self-Rating Anxiety Scale (SAS) and the Self-Rating Depression Scale (SDS). Each scale has several sub-item scores. The vectorized form of psychological assessment data is:

[0103]

[0104] in, is the eigenvector of psychological assessment data, for -dimensional feature vector; The first and second items of the anxiety self-rating scale or depression self-rating scale are respectively... Item rating.

[0105] (3) Wearable device data is usually time series data, recording changes in employees’ physiological characteristics (such as heart rate, sleep, number of steps, etc.). The sliding window method is used to convert time series data into feature vectors:

[0106]

[0107] in, is the M-dimensional time series feature vector of the wearable device data at the tth time step; They are the wearable device data collected at the t, t+1…t+M-1 time steps respectively.

[0108] In step S3, the feature vectors corresponding to the physical examination data and the psychological assessment data are respectively input into the trained multi-layer perceptron for feature extraction to obtain the physical examination feature vector and the assessment feature vector;

[0109] During implementation, the method for extracting features from the feature vectors of the physical examination data using a multi-layer perceptron is preferably as follows:

[0110] S31. Use the first connection layer of the multilayer perceptron to map the feature vector of the physical examination data to a high-dimensional space:

[0111]

[0112] in, It is a high-dimensional feature of physical examination data, with a dimension of 128; is the weight matrix of the first connection layer; is the bias term of the first connection layer; is the activation function, To obtain the maximum value function, is the feature vector of the physical examination data;

[0113] S32, use the second connection layer of the multi-layer perceptron to extract the high-dimensional features of the physical examination data to obtain a 64-dimensional physical examination feature vector :

[0114]

[0115] in, is the weight matrix of the second connection layer; is the bias term of the second connection layer;

[0116] S33, use the output layer of the multi-layer perceptron to output the physical examination feature vector ;

[0117] The method of extracting the physical examination feature vector and the evaluation feature vector by the multilayer perceptron is the same. The output dimension of the first connection layer of the evaluation feature vector is 256, and the output dimension of the second connection layer is 128.

[0118] In step S4, the feature vector of the wearable device data is input into the trained long short-term memory network for feature extraction to obtain a wear feature vector. The specific implementation method of this step includes:

[0119] S41. Calculate the output of the input gate of the long short-term memory network at the tth time step :

[0120]

[0121] in, 、 are the weight and bias of the input gate respectively; is the hidden state at the t-1th time step; is the M-dimensional time series feature vector of the wearable device data at the tth time step; M is the length of the time window; The sigmoid function limits the output to the range of [0,1]. is the feature vector of M dimension;

[0122] S42. Calculate the output of the forget gate of the long short-term memory network at the tth time step :

[0123]

[0124] in, and are the weight and bias of the forget gate respectively;

[0125] S43. Update the cell state at the tth time step according to the output of the input gate and the output of the forget gate:

[0126] ,

[0127] in, and are the cell states at the tth and t-1th time steps respectively; The parameters calculated from the input of the t-th time step and the hidden state of the previous time step; and They are The weights and biases of is the hyperbolic tangent function, which is an activation function that compresses the input value into the interval [−1,1].

[0128] S44. Calculate the output of the output gate of the long short-term memory network at the tth time step :

[0129]

[0130] in, and are the weight and bias of the output gate respectively;

[0131] S45. Calculate the features extracted by the long short-term memory network at the tth time step based on the output of the output gate and the unit state :

[0132]

[0133] S46, repeat steps S41 to S45 until t=M, and output the feature vector extracted at the Mth time step , as the wear feature vector.

[0134] In step S5, the physical examination feature vector, the assessment feature vector, and the wearing feature vector are input into the fusion module of the trained multimodal learning model for feature fusion, and the fused features are input into the DNN model of the multimodal learning model to generate a work ability score.

[0135] In one embodiment of the present invention, the method for the fusion module to perform feature fusion includes:

[0136] S51, perform feature splicing on the physical examination feature vector and the wearing feature vector to obtain the first splicing feature :

[0137]

[0138] Among them, Concat() is a concatenation function used to combine two or more feature vectors into one during machine learning or data processing; is the physical examination feature vector; is the wear feature vector;

[0139] S52, use the fully connected layer to the first splicing feature Perform dimensionality reduction operation to obtain the first dimensionality reduction data of 128 dimensions:

[0140]

[0141] in, is the first dimension reduction data; f() is the activation function Relu, which is used to perform nonlinear processing on the linearly transformed input in order to introduce the complexity and expression ability of the model; is the weight matrix of the physical examination data features, Perform linear changes; is the bias vector of the first splicing feature;

[0142] S53, perform feature splicing on the evaluation feature vector and the first dimension reduction data to obtain the global feature :

[0143]

[0144] in, is the evaluation feature vector;

[0145] S54, using the fully connected layer to analyze the global features Perform dimensionality reduction to obtain 64-dimensional fused features :

[0146]

[0147] in, is the feature weight matrix after modality fusion; is the bias vector of the global feature.

[0148] During implementation, the total loss function of the multimodal learning model is preferably expressed as:

[0149]

[0150] in, is the total loss function; is the weight of the i-th feature vector in step S5; L i is the loss value of the i-th feature vector in step S5; and are regularization coefficients; h i and h j are the i-th and j-th feature vectors in step S5 respectively; n is the total number of feature vectors input into the multimodal learning model for each employee in step S5; T is the transpose.

[0151] The loss function of this solution can dynamically adjust the weights based on the contribution of each modality to the prediction task, thereby improving the model's predictive performance. In addition, the regularization terms in the second and third terms on the right side of the loss function can effectively prevent a single modality from monopolizing too much weight, ensuring that the features of each modality are reasonably utilized in the model. At the same time, the modal diversity constraint (the first term on the right side of the loss function) further encourages the differences between different modal features, reduces feature redundancy, and improves the model's generalization ability. In other words, the loss function constructed in this solution can flexibly cope with complex multimodal data scenarios while ensuring feature balance, significantly improving the model's prediction effect on employee work ability.

[0152] In one embodiment of the present invention, the DNN model includes multiple layers of fully connected layers and output layers, and each layer uses a nonlinear activation function ReLU;

[0153] The first hidden layer contains 256 neurons, and its nonlinear activation function ReLU is:

[0154]

[0155] in, is the weight matrix of the first hidden layer, whose dimension is 256×64; is the bias vector of the first hidden layer, whose dimension is 256; is the output of the first hidden layer, whose dimension is 256;

[0156] The second hidden layer contains 128 neurons, and its nonlinear activation function ReLU is:

[0157]

[0158] in, is the weight matrix of the second hidden layer, whose dimension is 128×256; is the bias vector of the second hidden layer, whose dimension is 128; is the output of the second hidden layer, whose dimension is 128;

[0159] The third hidden layer contains 64 neurons, and its nonlinear activation function ReLU is:

[0160]

[0161] in, is the weight matrix of the third hidden layer, whose dimension is 64×128; is the bias vector of the third hidden layer, whose dimension is 64; is the output of the third hidden layer, whose dimension is 64;

[0162] The expression of the output layer is:

[0163]

[0164] in, is the weight matrix of the output layer, whose dimension is 1×64; is the bias vector of the output layer; Rate work ability.

[0165] In one embodiment of the present invention, a method for training a multimodal learning model includes:

[0166] Using the method of steps S1 to S4, obtaining the physical examination feature vectors, assessment feature vectors, and wearing feature vectors of several employees;

[0167] The method of steps S51 to S54 is used to obtain the fused features of each employee and obtain the work ability score of each employee;

[0168] The fused features of each employee are used as the model input features in the training set, and the work ability score of each employee is used as the model output feature in the training set;

[0169] The training set is input into the DNN model in batches, and the DNN model is trained using the adaptive modal weight optimization algorithm AMWO until the model converges or the number of training times reaches the preset number, obtaining a trained DNN model;

[0170] The fusion module and the trained DNN model are used to form a trained multimodal learning model.

[0171] When training the DNN model, the expression for updating the model parameters of the i-th feature vector is:

[0172] ,

[0173] in, and are the model parameters of the i-th feature vector in step S5 during the m+1-th and m-th training times, respectively; for gradient; is the learning rate of the i-th eigenvector in step S5; Hyperparameters adjusted to control the learning rate; is the initial learning rate; is the weight of the i-th feature vector in step S5; L i is the loss value of the i-th feature vector in step S5.

[0174] During implementation, the method for obtaining the work ability score preferably includes:

[0175] Obtain multiple health indicator values ​​from employees' physical examination data, scores from multiple self-assessment scales from psychological assessment data, and scores from multiple scoring criteria from work performance data; the multiple scoring criteria include performance indicators, attendance rate, work efficiency, teamwork and communication skills, and subjective scores from managers;

[0176] According to the scoring table corresponding to the physical examination data, the score of each health indicator value is obtained;

[0177] Based on the scores of health index values, self-assessment scales, and scoring criteria indicators, a weighted average method is used to obtain the health data score, mental health score, and work performance score:

[0178]

[0179]

[0180]

[0181] in, 、 and They are the indicator weights of BMI, blood pressure and blood sugar; and are the indicator weights of the anxiety self-assessment score and the depression self-assessment score respectively; 、 、 、 、 and These are the weights of performance score, attendance score, efficiency score, teamwork score and subjective score of managers.

[0182] Based on the health data score, mental health score and work performance score, a weighted average method is used to obtain the work ability score:

[0183]

[0184] in, 、 and These are the weights of health data score, mental health score and work performance score respectively.

[0185] In summary, the evaluation method of this solution generates relatively comprehensive health fusion data by integrating multi-source data (physical examination data, wearable device data, psychological assessment data, etc.). This data fusion goes beyond the limitations of traditional single-modal health management systems and can provide more accurate work ability assessments based on employees' health status, so that managers can accurately know whether employees are currently competent for their jobs and make timely adjustments to their jobs to ensure the physical and mental health of employees, reduce economic losses caused by employee health problems, and enhance employees' work enthusiasm and long-term work ability, ultimately improving the overall production efficiency of the enterprise.

Claims

1. An employee work ability assessment method based on multimodal health data fusion, characterized by: Including steps: S1. Obtain multiple health indicator values ​​from the employee's physical examination data, multiple self-assessment scale scores from the psychological assessment data, and multiple physiological monitoring indicator values ​​from the wearable device data; S2. Standardize the health index values, self-assessment scale scores, and physiological monitoring index values, and convert the standardized physical examination data, wearable device data, and psychological assessment data into feature vectors respectively; S3. Input the feature vectors corresponding to the physical examination data and the psychological assessment data into the trained multi-layer perceptron for feature extraction, thereby obtaining a physical examination feature vector and an assessment feature vector; S4. Input the feature vector of the wearable device data into the trained long short-term memory network for feature extraction to obtain a wear feature vector; S5. Input the physical examination feature vector, the assessment feature vector, and the wearing feature vector into the fusion module of the trained multimodal learning model to perform feature fusion, and input the fused features into the DNN model of the multimodal learning model to generate a work ability score; Training methods for multimodal learning models include: Obtain the physical examination feature vector, assessment feature vector, and wearing feature vector of several employees; Obtain the fused features of each employee and obtain the work ability score of each employee; The fused features of each employee are used as the model input features in the training set, and the work ability score of each employee is used as the model output feature in the training set; The training set is input into the DNN model in batches, and the DNN model is trained using the adaptive modal weight optimization algorithm AMWO until the model converges or the number of training times reaches the preset number, obtaining a trained DNN model; The fusion module and the trained DNN model are used to form a trained multimodal learning model; The total loss function of the multimodal learning model is expressed as: in, is the total loss function; is the weight of the i-th feature vector in step S5; L i is the loss value of the i-th feature vector in step S5; and are regularization coefficients; h i and h j are the i-th and j-th feature vectors in step S5, respectively; n is the total number of feature vectors input into the multimodal learning model for each employee in step S5; T is the transpose; When training the DNN model, the expression for updating the model parameters of the i-th feature vector is: ， in, and are the model parameters of the i-th feature vector in step S5 during the m+1-th and m-th training times, respectively; for gradient; is the learning rate of the i-th eigenvector in step S5; Hyperparameters adjusted to control the learning rate; is the initial learning rate; is the weight of the i-th feature vector in step S5; L i is the loss value of the i-th feature vector in step S5.

2. The employee work ability assessment method based on multimodal health data fusion according to claim 1 is characterized in that: The feature fusion methods of the fusion module include: S51, perform feature splicing on the physical examination feature vector and the wearing feature vector to obtain the first splicing feature : Among them, Concat() is the concatenation function; is the physical examination feature vector; is the wear feature vector; S52, use the fully connected layer to the first splicing feature Perform dimensionality reduction operation to obtain the first dimensionality reduction data of 128 dimensions: in, is the first dimension reduction data; f() is the activation function Relu; is the weight matrix of the physical examination data features; is the bias vector of the first splicing feature; S53, perform feature splicing on the evaluation feature vector and the first dimension reduction data to obtain the global feature : in, is the evaluation feature vector; S54, using the fully connected layer to analyze the global features Perform dimensionality reduction to obtain 64-dimensional fused features : in, is the feature weight matrix after modality fusion; is the bias vector of the global feature.

3. The employee work ability assessment method based on multimodal health data fusion according to claim 2 is characterized in that: The DNN model includes multiple layers of fully connected layers and output layers, and each layer uses a nonlinear activation function ReLU; The first hidden layer contains 256 neurons, and its nonlinear activation function ReLU is: in, is the weight matrix of the first hidden layer, whose dimension is 256×64; is the bias vector of the first hidden layer, whose dimension is 256; is the output of the first hidden layer, whose dimension is 256; The second hidden layer contains 128 neurons, and its nonlinear activation function ReLU is: in, is the weight matrix of the second hidden layer, whose dimension is 128×256; is the bias vector of the second hidden layer, whose dimension is 128; is the output of the second hidden layer, whose dimension is 128; The third hidden layer contains 64 neurons, and its nonlinear activation function ReLU is: in, is the weight matrix of the third hidden layer, whose dimension is 64×128; is the bias vector of the third hidden layer, whose dimension is 64; is the output of the third hidden layer, whose dimension is 64; The expression of the output layer is: in, is the weight matrix of the output layer, whose dimension is 1×64; is the bias vector of the output layer; Rate work ability.

4. The employee work ability assessment method based on multimodal health data fusion according to claim 2 is characterized in that: The method of extracting features from the feature vector of physical examination data using a multi-layer perceptron includes: S31. Use the first connection layer of the multilayer perceptron to map the feature vector of the physical examination data to a high-dimensional space: in, It is a high-dimensional feature of physical examination data, with a dimension of 128; is the weight matrix of the first connection layer; is the bias term of the first connection layer; is the activation function, To obtain the maximum value function, is the feature vector of the physical examination data; S32, use the second connection layer of the multi-layer perceptron to extract the high-dimensional features of the physical examination data to obtain a 64-dimensional physical examination feature vector : in, is the weight matrix of the second connection layer; is the bias term of the second connection layer; S33, use the output layer of the multi-layer perceptron to output the physical examination feature vector ; The method of extracting the physical examination feature vector and the evaluation feature vector by the multi-layer perceptron is the same. The output dimension of the first connection layer of the evaluation feature vector is 256, and the output dimension of the second connection layer is 128. The method of extracting features from the feature vectors of wearable device data using a long short-term memory network includes: S41. Calculate the output of the input gate of the long short-term memory network at the tth time step : in, 、 are the weight and bias of the input gate respectively; is the hidden state at the t-1th time step; is the M-dimensional time series feature vector of the wearable device data at the tth time step; M is the length of the time window; is the sigmoid function; is the feature vector of M dimension; S42. Calculate the output of the forget gate of the long short-term memory network at the tth time step : in, and are the weight and bias of the forget gate respectively; S43. Update the cell state at the tth time step according to the output of the input gate and the output of the forget gate: ， in, and are the cell states at the tth and t-1th time steps respectively; The parameters calculated from the input of the t-th time step and the hidden state of the previous time step; and They are The weights and biases of is the hyperbolic tangent function; S44. Calculate the output of the output gate of the long short-term memory network at the tth time step : in, and are the weight and bias of the output gate respectively; S45. Calculate the features extracted by the long short-term memory network at the tth time step based on the output of the output gate and the unit state : S46, repeat steps S41 to S45 until t=M, and output the feature vector extracted at the Mth time step , as the wearing feature vector.

5. The employee work ability assessment method based on multimodal health data fusion according to claim 4 is characterized in that: Training methods for multimodal learning models include: Using the method of steps S1 to S4, obtaining the physical examination feature vectors, assessment feature vectors, and wearing feature vectors of several employees; The method of steps S51 to S54 is used to obtain the fused features of each employee and obtain the work ability score of each employee.

6. The employee work ability assessment method based on multimodal health data fusion according to claim 5 is characterized in that: The method for obtaining the work ability score includes: Obtain multiple health indicator values ​​from employees' physical examination data, scores from multiple self-assessment scales from psychological assessment data, and scores from multiple scoring standard indicators from work performance data; According to the scoring table corresponding to the physical examination data, the score of each health indicator value is obtained; Based on the scores of health index values, self-assessment scales, and scoring criteria indicators, a weighted average method is used to obtain health data scores, mental health scores, and work performance scores; The work ability score is obtained by taking a weighted average approach based on the health data score, mental health score and work performance score.

7. The employee work ability assessment method based on multimodal health data fusion according to claim 6 is characterized in that: Multiple health indicators include BMI, blood pressure and blood sugar; multiple self-assessment scales include anxiety self-assessment scale and depression self-assessment scale; multiple scoring standard indicators include performance indicators, attendance rate, work efficiency, teamwork and communication skills and subjective scores of managers; multiple physiological monitoring indicators include heart rate, sleep and activity level.

Images