A workload prediction system for internal and external surgery nursing based on big data

Abstract

The present invention discloses an internal and external surgical nursing workload prediction system based on big data, including a data acquisition module, a data preprocessing module, a multidimensional workload factor extraction module, a workload prediction model establishment module, a model training optimization module and an internal and external surgical nursing work management module. The present invention relates to the field of nursing work information data processing technology, and specifically refers to an internal and external surgical nursing workload prediction system based on big data. This solution innovatively proposes a dynamic feature weight adjustment mechanism and spatiotemporal feature fusion to effectively capture the spatiotemporal correlation of the nursing workload; innovatively proposes a multi-level prediction architecture that can comprehensively capture the relationship between each level, thereby improving the accuracy of the entire nursing workload prediction result; innovatively enhances the training optimization effect of the model through a cross-level error compensation mechanism and an improved optimization algorithm, and significantly improves the application performance in nursing workload prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of nursing work information data processing, and in particular to an internal and external surgery nursing workload prediction system based on big data. Background Art

[0002] With the continuous growth of demand for medical services, hospitals, especially internal and external medicine nursing departments, are facing tremendous work pressure. The workload of nursing staff directly affects the quality of nursing services, patient safety and the overall operational efficiency of the hospital. In order to meet these challenges, an internal and external medicine nursing workload prediction system based on big data has emerged. By collecting, analyzing and processing massive data related to internal and external medicine nursing work in hospitals, advanced machine learning and big data technologies are used to predict the workload of nursing staff and provide data support for hospital managers. However, the traditional nursing workload prediction system uses static weights and fixed feature selection when processing features, and cannot fully utilize the correlation between time and space to optimize the prediction results, resulting in inaccurate nursing workload prediction results. The existing nursing workload prediction system only focuses on a certain level of nursing workload prediction in traditional nursing workload prediction, ignoring the technical problem of mutual influence between different levels. In the traditional multi-level model training optimization, there are inappropriate model weights and built-in parameter settings, resulting in inaccurate nursing workload prediction results. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an internal and external surgical nursing workload prediction system based on big data. In view of the technical problems that the traditional nursing workload prediction system adopts static weights and fixed feature selection when processing features, and cannot make full use of the correlation between time and space to optimize the prediction results, thereby resulting in inaccurate nursing workload prediction, this solution innovatively proposes a dynamic feature weight adjustment mechanism and spatiotemporal feature fusion, and uses a multi-head attention mechanism to dynamically adjust the weights of features to ensure that the most relevant features can get higher weights in different nursing scenarios. By fusing the spatial workload factor and the temporal workload factor to generate the spatiotemporal workload factor, the spatiotemporal correlation of the nursing workload is effectively captured, the problem of spatiotemporal feature separation in the traditional method is solved, and the accuracy and reliability of the nursing workload prediction are significantly improved; in view of the technical problem that the existing nursing workload prediction only focuses on a certain level in the traditional nursing workload prediction and ignores the mutual influence between different levels, this solution innovatively proposes A multi-level prediction architecture is proposed. By constructing a three-level prediction architecture of department-level macro prediction layer, ward-level meso prediction layer and bed-level micro prediction layer, the relationship and dependency between each level can be fully captured, which solves the defect of ignoring the mutual influence between levels in traditional methods. The macro trend at department level, resource allocation at ward level and individual demand at bed level can all be predicted, thereby significantly improving the accuracy of the entire nursing workload prediction results. Aiming at the technical problem that the model weights and built-in parameters are inappropriately set in the traditional multi-level model training optimization, which leads to inaccurate nursing workload prediction results, this scheme ensures that the errors between different levels can be effectively transmitted and compensated through a cross-level error compensation mechanism and an improved optimization algorithm. The improved optimization algorithm can enhance the learning ability of each level, avoid the performance bottleneck caused by unreasonable parameter settings in traditional methods, and enhance the training optimization effect of the model, so that the nursing workload prediction results are more accurate, which can better meet the actual needs and significantly improve the application performance in nursing workload prediction.

[0004] The technical solution adopted by the present invention is as follows: the present invention provides a big data-based internal and external surgery nursing workload prediction system, including a data acquisition module, a data preprocessing module, a multi-dimensional workload factor extraction module, a workload prediction model establishment module, a model training optimization module and an internal and external surgery nursing work management module;

[0005] The data collection module obtains original data of nursing workload prediction by collecting data from the hospital management system;

[0006] The data preprocessing module specifically performs data cleaning, data noise filtering, spatiotemporal data alignment and data standardization on the raw data of the nursing workload prediction to obtain preliminary data of the nursing workload prediction;

[0007] The multi-dimensional workload factor extraction module is used to extract multi-dimensional workload factors by constructing a workload factor pool and screening out key workload factors therefrom, generating spatiotemporal workload factors at the same time, using a multi-head attention mechanism to dynamically obtain the weights of the workload factors, and merging the weights with each workload factor to obtain an optimized workload factor combination;

[0008] The workload prediction model establishment module constructs a three-level prediction architecture of a department-level macro prediction layer, a ward-level meso prediction layer, and a bed-level micro prediction layer, integrates the prediction results of the department-level macro prediction layer, the ward-level meso prediction layer, and the bed-level micro prediction layer, generates a comprehensive multi-level nursing workload prediction result, and obtains a workload prediction model;

[0009] The model training optimization module optimizes the training process of the workload prediction model by introducing a cross-level error compensation mechanism, and uses an improved optimization algorithm to adjust the workload prediction model parameters to obtain an optimal workload prediction model;

[0010] The internal and external medicine nursing work management module realizes nursing workload prediction based on the optimal workload prediction model and obtains multi-level nursing workload prediction results. Based on this data result, hospital managers can perform overall nursing work management and resource allocation.

[0011] Furthermore, the data collection module specifically obtains raw data for nursing workload prediction from the hospital management system by collecting data; the raw data for nursing workload prediction includes historical nursing workload data and real-time nursing workload data; the historical nursing workload data and real-time nursing workload data both include patient data, nursing staff data and department data; the historical nursing workload data also includes historical nursing workload results.

[0012] Furthermore, the data preprocessing module specifically performs data cleaning, data noise filtering, spatiotemporal data alignment and data standardization on the raw data of the nursing workload prediction to obtain preliminary data of the nursing workload prediction;

[0013] The data cleaning is used to ensure the consistency, accuracy and completeness of the original data, specifically to remove duplicate data, correct data format errors, process missing data values, and delete inconsistent and unrepairable data;

[0014] The data noise filtering is used to identify and process abnormal fluctuations and erroneous values ​​in the data, specifically using a sliding window algorithm and context information to perform dynamic threshold detection on the data, identify and remove abnormal data and erroneous values;

[0015] The spatiotemporal data alignment is used to ensure the consistency of data from multiple sources in the same time frame and spatial dimension. Specifically, the dynamic time warping algorithm is used to calculate the best match of the time series, process data sources with inconsistent timestamps, ensure the time series data is time-aligned, and spatially align the data from different departments, wards and beds by constructing a three-level spatial mapping model; the three-level spatial mapping model is department-ward-bed;

[0016] The data standardization is used to enable data of different dimensions to be compared on the same scale, ensuring that each dimension of the data is within the same dimensional range, and specifically is to perform data standardization processing based on the maximum and minimum normalization method.

[0017] Furthermore, the multi-dimensional workload factor refining module specifically includes the following steps:

[0018] Construct a workload factor pool, specifically extract multidimensional workload factors from preliminary data of nursing workload prediction, and integrate them into a workload factor pool;

[0019] Screening key workload factors, which is used to screen out workload factors with high impact on nursing workload prediction from the workload factor pool, specifically using the mutual information algorithm to calculate the correlation between the workload factor and the nursing workload, obtaining the workload factor with a mutual information value higher than a threshold, and obtaining the key workload factor;

[0020] Generate a spatiotemporal workload factor to optimize the combination of spatial and temporal workload factors and capture the spatiotemporal correlation of nursing work. Specifically, extract the spatial workload factor and the temporal workload factor, and combine the two to generate a new spatiotemporal workload factor, including the following steps:

[0021] Spatial workload factor extraction, based on the spatial relationship of the three-level spatial mapping model, a graph embedding algorithm is used to generate spatial workload factors to capture the nursing synergy between adjacent beds;

[0022] Temporal workload factor extraction, using wavelet transform to extract multi-scale temporal workload factors and capture the temporal variation trend in the nursing process;

[0023] To obtain the spatiotemporal workload factor, the formula used is as follows:

[0024] ;

[0025] In the formula, represents the jth spatiotemporal workload factor, represents the weight of the spatial workload factor, represents the weight of the time workload factor, represents the spatial workload factor, represents the time workload factor;

[0026] Obtain dynamic workload factor weights so that the most relevant features can obtain higher weights in different nursing scenarios. Specifically, the weights of each key workload factor and spatiotemporal workload factor are dynamically adjusted using a multi-head attention mechanism.

[0027] Generate an optimized workload factor combination, specifically by combining the dynamically adjusted workload factor weights with the key workload factor and the spatiotemporal workload factor to generate an optimized workload factor combination, and the formula used is as follows:

[0028] ;

[0029] In the formula, represents the optimized workload factor combination, represents the i-th critical workload factor, represents the jth spatiotemporal workload factor, represents the dynamically adjusted weight of the i-th critical workload factor, represents the dynamically adjusted weight of the jth spatiotemporal workload factor, represents the feature cross-weight parameter, Represents the Sigmoid activation function.

[0030] Furthermore, the module for establishing a workload prediction model is used to establish a workload prediction model to predict nursing workloads of different levels, thereby obtaining a workload prediction model; specifically, the following steps are included:

[0031] Construct a department-level macro prediction layer to predict the nursing workload of the entire department. Specifically, this layer uses a bidirectional long short-term memory neural network as the prediction network of the department-level macro prediction layer to obtain the department-level workload results for the next 24 hours as the input data of the ward-level meso prediction layer.

[0032] A ward-level meso-level prediction layer is constructed to predict the nursing workload of each ward. It can capture the relationship between resource usage, number of patients and nursing hours in the ward. Specifically, this layer uses a recursive neural network as the prediction network of the ward-level meso-level prediction layer to obtain the nursing workload results of the ward level in the next 12 hours as the input data of the bed-level micro-prediction layer.

[0033] Construct a bed-level micro-prediction layer to predict the nursing load based on individualized bed demand and patient condition. Specifically, this layer uses a graph neural network as the prediction network of the bed-level micro-prediction layer to obtain the bed-level nursing workload results for the next 4 hours.

[0034] Generate a multi-level nursing workload prediction result, specifically by integrating the prediction results of the department-level macro prediction layer, the ward-level meso prediction layer, and the bed-level micro prediction layer to generate a comprehensive multi-level nursing workload prediction result.

[0035] Furthermore, the model training optimization module is used to train and optimize the model, improve the prediction results of the model, and obtain the optimal workload prediction model; specifically, the following steps are included:

[0036] Workload prediction model training is used to ensure that the model learns and captures the patterns and trends of nursing workload. Specifically, it includes the following steps:

[0037] Prepare training data, specifically use the historical nursing workload data in the preliminary data for nursing workload prediction as training data, randomly divide the training data into three equal parts, one for training the department-level macro prediction layer, another for training the ward-level meso prediction layer, and the remaining part for training the bed-level micro prediction layer;

[0038] A cross-level error compensation mechanism is introduced. Specifically, in the training process of each level, the prediction error of the current level is first calculated. Through the cross-level error compensation mechanism, the error of the upstream level is passed to the downstream level, and the error of the upstream level is used as an additional constraint term of the loss function of the downstream level. The loss function is minimized through the back propagation algorithm, the weight of each level is adjusted, and the prediction effect of the model is optimized; the following steps are included:

[0039] Department-level layer training, specifically, calculates the prediction error value of this layer, updates the weight of the department-level macro prediction layer through the back propagation algorithm, and transfers it to the ward-level meso prediction layer; the formula used is as follows:

[0040] ;

[0041] In the formula, represents the error value of the department-level macro-forecast layer, represents the total number of department-level macro prediction layer training data, represents the true value of the i-th sample, represents the predicted value of the department-level macro-forecast layer of the i-th sample;

[0042] Ward-level layer training involves calculating the prediction error value of this layer, updating the weights of the ward-level meso-prediction layer through the back-propagation algorithm, and transferring the error to the bed-level micro-prediction layer. The formula used is as follows:

[0043] ;

[0044] In the formula, represents the error value of the ward-level meso-prediction layer, represents the total number of training data for the ward-level meso-prediction layer, represents the true value of the jth sample, represents the predicted value of the ward-level meso-level prediction layer of the jth sample, represents the weighting coefficient of the error of the macro prediction layer at the department level;

[0045] Bed-level layer training, specifically, calculates the prediction error of this layer and updates the weights of the bed-level micro-prediction layer through the back-propagation algorithm; the formula used is as follows:

[0046] ;

[0047] In the formula, represents the error value of the bed-level micro-prediction layer, represents the total number of training data for the bed-level micro prediction layer, represents the true value of the kth sample, represents the predicted value of the bed-level micro-prediction layer of the k-th sample, represents the weighting coefficient of the ward-level meso-prediction layer error;

[0048] The workload prediction model parameter optimization is specifically to use an improved optimization algorithm to adjust the parameters of each level of the workload prediction model respectively, and specifically includes the following steps;

[0049] Initialization parameters, specifically, by constructing the initial parameters of the algorithm; the initial parameters of the algorithm include the number of individuals N and the maximum number of iterations ;

[0050] Initialize the population, specifically, the individuals in the population have an initial position and initial velocity in the solution space, and the initial position and initial velocity are obtained randomly;

[0051] Population individual fitness evaluation, specifically calculating the individual fitness value F in the population i ; The performance of each prediction layer of the workload prediction model established based on the individual position is used as the fitness value of the individual;

[0052] Grouping of individuals in the population, specifically using a collaborative grouping optimization strategy to divide the population into multiple groups. Each group has a leader, whose position is the best position in the group. Individuals in each group search around the leader.

[0053] Get individual search parameters using the following formula:

[0054] ;

[0055] In the formula, Represents the step size factor, which is used to adjust the balance between global search and local search. represents the frequency of the global best individual, Indicates the current iteration number, represents the movement frequency of the i-th individual, represents the minimum frequency of individual movement, represents the maximum frequency of individual movement, represents a random number uniformly distributed in the range [0,1]. represents the search balance factor, Represents a tuning parameter that affects the decay rate of the search balance factor;

[0056] Update the individual properties of the population, specifically control the search phase according to the search balance factor, the formula used is as follows:

[0057] ;

[0058] In the formula, represents the position of the i-th individual in the t+1-th iteration, represents the position of the i-th individual in the t-th iteration, represents the leader of the individual in the jth group, represents the position of the rth individual in the tth iteration, represents a random variable drawn from a normal distribution, represents a random variable drawn from a standard normal distribution, represents the movement speed of the i-th individual in the t+1-th iteration, represents a random number uniformly distributed in the range [0,1]. represents a random number uniformly distributed in the range [0.5,1], Represents a random number uniformly distributed in the range [-1,1], represents the search intensity of the i-th individual in the t-th iteration, represents the search capability of the i-th individual in the t-th iteration;

[0059] End the current iteration, specifically by comparing the fitness values ​​to obtain the global optimal position, and replace the group with the worst fitness value in the group in the population with the average position of the remaining groups of individuals, and update the search intensity parameter and search ability parameter at the same time; the formula used is as follows:

[0060] ;

[0061] ;

[0062] In the formula, represents the fitness function, represents the global optimal position of an individual;

[0063] The optimal solution is obtained by i When the fitness threshold is higher than the set fitness threshold and the maximum number of iterations is reached, the search is terminated and the individual global optimal position is obtained, where the individual global optimal position specifically refers to the optimal parameters of each level of the workload prediction model;

[0064] The optimal workload prediction model is obtained, specifically, the optimal workload prediction model is obtained through workload prediction model training and workload prediction model parameter optimization.

[0065] Furthermore, the internal and external medicine nursing work management module specifically inputs the real-time nursing workload data of the preliminary nursing workload prediction data into the multi-dimensional workload factor refining module, refines and optimizes the workload factor through the module, obtains the optimized workload factor combination, and inputs the optimized workload factor combination into the optimal workload prediction model to obtain multi-level nursing workload prediction results. Based on the multi-level nursing workload prediction results, hospital managers can perform accurate nursing resource allocation, scheduling optimization and nursing workload balance to improve the overall nursing management efficiency.

[0066] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0067] (1) In response to the technical problems that traditional nursing workload prediction systems use static weights and fixed feature selection when processing features, and cannot make full use of the temporal and spatial correlations to optimize the prediction results, resulting in inaccurate nursing workload prediction, this scheme innovatively proposes a dynamic feature weight adjustment mechanism and spatiotemporal feature fusion, and uses a multi-head attention mechanism to dynamically adjust the feature weights to ensure that the most relevant features can get higher weights in different nursing scenarios. By fusing the spatial workload factor and the temporal workload factor to generate the spatiotemporal workload factor, the spatiotemporal correlation of the nursing workload is effectively captured, which solves the problem of spatiotemporal feature separation in traditional methods and significantly improves the accuracy and reliability of nursing workload prediction.

[0068] (2) In response to the technical problem that the existing traditional nursing workload forecasting only focuses on a certain level and ignores the mutual influence between different levels, this solution innovatively proposes a multi-level forecasting architecture. By constructing a three-level forecasting architecture consisting of a department-level macro-forecasting layer, a ward-level meso-forecasting layer, and a bed-level micro-forecasting layer, it can comprehensively capture the relationship and dependency between the levels and solve the defect of ignoring the mutual influence between the levels in the traditional method. The macro trend at the department level, the resource allocation at the ward level, and the individual needs at the bed level can all be predicted, thereby significantly improving the accuracy of the overall nursing workload forecasting results.

[0069] (3) In order to address the technical problem that inappropriate model weights and built-in parameter settings exist in traditional multi-level model training optimization, which leads to inaccurate nursing workload prediction results, this solution uses a cross-level error compensation mechanism and an improved optimization algorithm to ensure that errors between different levels can be effectively transmitted and compensated. The improved optimization algorithm can enhance the learning ability of each level, avoid the performance bottleneck caused by unreasonable parameter settings in traditional methods, and enhance the training optimization effect of the model, so that the nursing workload prediction results are more accurate and can better meet actual needs, and significantly improve the application performance in nursing workload prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A module schematic diagram of a big data-based internal and external surgery nursing workload prediction system provided by the present invention;

[0071] Figure 2 It is a flowchart of the multi-dimensional workload factor extraction module;

[0072] Figure 3 A flowchart for establishing a workload prediction model module;

[0073] Figure 4 Schematic diagram of the process of model training optimization module;

[0074] Figure 5 A flowchart of the cross-level error compensation mechanism introduced into the model training optimization module;

[0075] Figure 6 A schematic diagram of the process of optimizing the workload prediction model parameters in the model training optimization module;

[0076] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0077] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0078] In the description of the present invention, it is necessary to understand that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred system or element must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0079] Example 1, see Figure 1 The present invention provides a big data-based internal and external surgery nursing workload prediction system, which includes a data acquisition module, a data preprocessing module, a multi-dimensional workload factor extraction module, a workload prediction model establishment module, a model training optimization module and an internal and external surgery nursing work management module;

[0080] The data acquisition module acquires raw data of nursing workload prediction from the hospital management system by collecting data, and sends the data to the data preprocessing module;

[0081] The data preprocessing module receives the data sent by the data acquisition module, specifically performs data cleaning, data noise filtering, spatiotemporal data alignment and data standardization on the raw data of the nursing workload prediction, obtains preliminary data of the nursing workload prediction, and sends the data to the multidimensional workload factor refining module;

[0082] The multi-dimensional workload factor extraction module receives the data sent by the data preprocessing module, constructs a workload factor pool, screens out key workload factors therefrom, generates spatiotemporal workload factors, uses a multi-head attention mechanism to dynamically obtain the weights of the workload factors, merges the weights with each workload factor, thereby obtaining an optimized workload factor combination, and sends the data to the internal and external medicine nursing work management module;

[0083] The workload prediction model establishment module receives the data sent by the multi-dimensional workload factor extraction module, and integrates the prediction results of the department-level macro prediction layer, the ward-level meso prediction layer, and the bed-level micro prediction layer by constructing a three-level prediction architecture of the department-level macro prediction layer, the ward-level meso prediction layer, and the bed-level micro prediction layer to generate a comprehensive multi-level nursing workload prediction result, obtain a workload prediction model, and send the data to the model training optimization module;

[0084] The model training optimization module receives data sent by the workload prediction model establishment module, optimizes the training process of the workload prediction model by introducing a cross-level error compensation mechanism, and simultaneously uses an improved optimization algorithm to adjust the workload prediction model parameters to obtain an optimal workload prediction model;

[0085] The internal and external medicine nursing work management module receives data sent by the model training optimization module, implements nursing workload prediction based on the optimal workload prediction model, and obtains multi-level nursing workload prediction results. Based on this data result, hospital managers can perform overall nursing work management and resource allocation.

[0086] Example 2, see Figure 1, this embodiment is based on the above embodiment, and the data acquisition module specifically obtains the original data of nursing workload prediction from the hospital management system by acquisition; the original data of nursing workload prediction includes historical nursing workload data and real-time nursing workload data; the historical nursing workload data and real-time nursing workload data both include patient data, nursing staff data and department data; the historical nursing workload data also includes historical nursing workload results; the patient data includes patient basic information, patient disease type, patient condition assessment and patient nursing needs; the patient nursing needs are specifically wound care, drug infusion and patient monitoring needs; the nursing staff data includes nursing personal information data, nursing personal work arrangements, nursing skills and nursing experience; the nursing personal information data includes name, title and qualification certification; the department data includes department type, department capacity, department equipment type, department equipment status, the ratio of nurses to patients in the department, department ward information data and department bed information data; the department ward information data includes the number of patients in the ward, the number of nursing staff configured in the ward and the medical equipment in the ward.

[0087] Example 3, see Figure 1 , this embodiment is based on the above embodiment, and the data preprocessing module specifically performs data cleaning, data noise filtering, spatiotemporal data alignment and data standardization on the raw data of the nursing workload prediction to obtain preliminary data of the nursing workload prediction;

[0088] The data cleaning is used to ensure the consistency, accuracy and completeness of the original data, specifically to remove duplicate data, correct data format errors, process missing data values, and delete inconsistent and unrepairable data;

[0089] The data noise filtering is used to identify and process abnormal fluctuations and erroneous values ​​in the data, specifically using a sliding window algorithm and context information to perform dynamic threshold detection on the data, identify and remove abnormal data and erroneous values;

[0090] The spatiotemporal data alignment is used to ensure the consistency of data from multiple sources in the same time frame and spatial dimension. Specifically, the dynamic time warping method is used to calculate the best match of the time series, process data sources with inconsistent timestamps, ensure the time series data is time-aligned, and spatially align the data from different departments, wards and beds by constructing a three-level spatial mapping model; the three-level spatial mapping model is department-ward-bed;

[0091] The data standardization is used to enable data of different dimensions to be compared on the same scale, ensuring that each dimension of the data is within the same dimensional range, and specifically is to perform data standardization processing based on the maximum and minimum normalization method.

[0092] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the multi-dimensional workload factor extraction module specifically includes the following steps:

[0093] Construct a workload factor pool, specifically extract multidimensional workload factors from preliminary data of nursing workload prediction, and integrate them into a workload factor pool;

[0094] Screening key workload factors, used to screen out workload factors with high impact on nursing workload prediction from the workload factor pool, specifically using a mutual information algorithm to calculate the correlation between the workload factor and the nursing workload, obtaining a workload factor with a mutual information value higher than a threshold, and obtaining a key workload factor; the specific value of the threshold is 0.4;

[0095] Generate a spatiotemporal workload factor to optimize the combination of spatial and temporal workload factors and capture the spatiotemporal correlation of nursing work. Specifically, extract the spatial workload factor and the temporal workload factor, and combine the two to generate a new spatiotemporal workload factor, including the following steps:

[0096] Spatial workload factor extraction, based on the spatial relationship of the three-level spatial mapping model, a graph embedding algorithm is used to generate spatial workload factors to capture the nursing synergy between adjacent beds;

[0097] Temporal workload factor extraction, using wavelet transform to extract multi-scale temporal workload factors and capture the temporal variation trend in the nursing process;

[0098] To obtain the spatiotemporal workload factor, the formula used is as follows:

[0099] ;

[0100] In the formula, represents the jth spatiotemporal workload factor, represents the weight of the spatial workload factor, represents the weight of the time workload factor, represents the spatial workload factor, represents the time workload factor;

[0101] Obtain dynamic workload factor weights so that the most relevant features can obtain higher weights in different nursing scenarios. Specifically, the weights of each key workload factor and spatiotemporal workload factor are dynamically adjusted using a multi-head attention mechanism.

[0102] Generate an optimized workload factor combination, specifically by combining the dynamically adjusted workload factor weights with the key workload factor and the spatiotemporal workload factor to generate an optimized workload factor combination, and the formula used is as follows:

[0103] ;

[0104] In the formula, represents the optimized workload factor combination, represents the i-th critical workload factor, represents the jth spatiotemporal workload factor, represents the dynamically adjusted weight of the i-th critical workload factor, represents the dynamically adjusted weight of the jth spatiotemporal workload factor, represents the feature cross-weight parameter, Represents the Sigmoid activation function.

[0105] By performing the above operations, in order to address the technical problems that the traditional nursing workload prediction system uses static weights and fixed feature selection when processing features, and cannot make full use of the correlation between time and space to optimize the prediction results, resulting in inaccurate nursing workload prediction, this solution innovatively proposes a dynamic feature weight adjustment mechanism and spatiotemporal feature fusion, and uses a multi-head attention mechanism to dynamically adjust the weights of features to ensure that the most relevant features can get higher weights in different nursing scenarios. By fusing the spatial workload factor and the temporal workload factor to generate the spatiotemporal workload factor, the spatiotemporal correlation of the nursing workload is effectively captured, the problem of spatiotemporal feature separation in the traditional method is solved, and the accuracy and reliability of the nursing workload prediction are significantly improved.

[0106] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the workload prediction model establishment module is used to establish a workload prediction model to predict nursing workloads of different levels to obtain a workload prediction model; specifically, the following steps are included:

[0107] Construct a department-level macro prediction layer to predict the nursing workload of the entire department. Specifically, this layer uses a bidirectional long short-term memory neural network as the prediction network of the department-level macro prediction layer to obtain the department-level workload results for the next 24 hours as the input data of the ward-level meso prediction layer.

[0108] A ward-level meso-level prediction layer is constructed to predict the nursing workload of each ward. It can capture the relationship between resource usage, number of patients and nursing hours in the ward. Specifically, this layer uses a recursive neural network as the prediction network of the ward-level meso-level prediction layer to obtain the nursing workload results of the ward level in the next 12 hours as the input data of the bed-level micro-prediction layer.

[0109] Construct a bed-level micro-prediction layer to predict the nursing load based on individualized bed demand and patient condition. Specifically, this layer uses a graph neural network as the prediction network of the bed-level micro-prediction layer to obtain the bed-level nursing workload results for the next 4 hours.

[0110] Generate a multi-level nursing workload prediction result, specifically by integrating the prediction results of the department-level macro prediction layer, the ward-level meso prediction layer, and the bed-level micro prediction layer to generate a comprehensive multi-level nursing workload prediction result.

[0111] By performing the above operations, this solution innovatively proposes a multi-level prediction architecture to address the technical problem that the existing traditional nursing workload prediction only focuses on a certain level and ignores the mutual influence between different levels. By constructing a three-level prediction architecture of department-level macro prediction layer, ward-level meso prediction layer and bed-level micro prediction layer, it can comprehensively capture the relationship and dependency between each level, and solve the defect of ignoring the mutual influence between levels in the traditional method. The macro trend at the department level, the resource allocation at the ward level, and the individual needs at the bed level can all be predicted, thereby significantly improving the accuracy of the entire nursing workload prediction results.

[0112] Example 6, see Figure 1 , Figure 4 , Figure 5 and Figure 6 This embodiment is based on the above embodiment, and the model training optimization module is used to train and optimize the model, improve the prediction results of the model, and obtain the optimal workload prediction model; specifically, it includes the following steps:

[0113] Workload prediction model training is used to ensure that the model learns and captures the patterns and trends of nursing workload. Specifically, it includes the following steps:

[0114] Prepare training data, specifically use the historical nursing workload data in the preliminary data for nursing workload prediction as training data, randomly divide the training data into three equal parts, one for training the department-level macro prediction layer, another for training the ward-level meso prediction layer, and the remaining part for training the bed-level micro prediction layer;

[0115] A cross-level error compensation mechanism is introduced. Specifically, in the training process of each level, the prediction error of the current level is first calculated. Through the cross-level error compensation mechanism, the error of the upstream level is passed to the downstream level, and the error of the upstream level is used as an additional constraint term of the loss function of the downstream level. The loss function is minimized through the back propagation algorithm, the weight of each level is adjusted, and the prediction effect of the model is optimized; the following steps are included:

[0116] Department-level layer training, specifically, calculates the prediction error value of this layer, updates the weight of the department-level macro prediction layer through the back propagation algorithm, and transfers it to the ward-level meso prediction layer; the formula used is as follows:

[0117] ;

[0118] In the formula, represents the error value of the department-level macro-forecast layer, represents the total number of department-level macro prediction layer training data, represents the true value of the i-th sample, represents the predicted value of the department-level macro-forecast layer of the i-th sample;

[0119] Ward-level layer training involves calculating the prediction error value of this layer, updating the weights of the ward-level meso-prediction layer through the back-propagation algorithm, and transferring the error to the bed-level micro-prediction layer. The formula used is as follows:

[0120] ;

[0121] In the formula, represents the error value of the ward-level meso-prediction layer, represents the total number of training data for the ward-level meso-prediction layer, represents the true value of the jth sample, represents the predicted value of the ward-level meso-level prediction layer of the jth sample, represents the weighting coefficient of the error of the macro prediction layer at the department level;

[0122] Bed-level layer training, specifically, calculates the prediction error of this layer and updates the weights of the bed-level micro-prediction layer through the back-propagation algorithm; the formula used is as follows:

[0123] ;

[0124] In the formula, represents the error value of the bed-level micro-prediction layer, represents the total number of training data for the bed-level micro prediction layer, represents the true value of the kth sample, represents the predicted value of the bed-level micro-prediction layer of the k-th sample, represents the weighting coefficient of the ward-level meso-prediction layer error;

[0125] The workload prediction model parameter optimization is specifically to use an improved optimization algorithm to adjust the parameters of each level of the workload prediction model respectively, and specifically includes the following steps;

[0126] Initialization parameters, specifically, by constructing the initial parameters of the algorithm; the initial parameters of the algorithm include the number of individuals N and the maximum number of iterations ;

[0127] Initialize the population, specifically, the individuals in the population have an initial position and initial velocity in the solution space, and the initial position and initial velocity are obtained randomly;

[0128] Population individual fitness evaluation, specifically calculating the individual fitness value F in the population i ; The performance of each prediction layer of the workload prediction model established based on the individual position is used as the fitness value of the individual;

[0129] Grouping of individuals in the population, specifically using a collaborative grouping optimization strategy to divide the population into multiple groups. Each group has a leader, whose position is the best position in the group. Individuals in each group search around the leader.

[0130] Get individual search parameters using the following formula:

[0131] ;

[0132] In the formula, Represents the step size factor, which is used to adjust the balance between global search and local search. represents the frequency of the global best individual, Indicates the current iteration number, represents the movement frequency of the i-th individual, represents the minimum frequency of individual movement, represents the maximum frequency of individual movement, represents a random number uniformly distributed in the range [0,1]. represents the search balance factor, Represents a tuning parameter that affects the decay rate of the search balance factor;

[0133] Update the individual properties of the population, specifically control the search phase according to the search balance factor, the formula used is as follows:

[0134] ;

[0135] In the formula, represents the position of the i-th individual in the t+1-th iteration, represents the position of the i-th individual in the t-th iteration, represents the leader of the individual in the jth group, represents the position of the rth individual in the tth iteration, represents a random variable drawn from a normal distribution, represents a random variable drawn from a standard normal distribution, represents the movement speed of the i-th individual in the t+1-th iteration, represents a random number uniformly distributed in the range [0,1]. represents a random number uniformly distributed in the range [0.5,1], Represents a random number uniformly distributed in the range [-1,1], represents the search intensity of the i-th individual in the t-th iteration, represents the search capability of the i-th individual in the t-th iteration;

[0136] End the current iteration, specifically by comparing the fitness values ​​to obtain the global optimal position, and replace the group with the worst fitness value in the group in the population with the average position of the remaining groups of individuals, and update the search intensity parameter and search ability parameter at the same time; the formula used is as follows:

[0137] ;

[0138] ;

[0139] ;

[0140] ;

[0141] In the formula, represents the fitness function, represents the global optimal position of an individual, represents the search intensity of the i-th individual in the t+1-th iteration, represents the search capability of the i-th individual in the t+1-th iteration, It represents the exploration capability enhancement coefficient, which is a constant;

[0142] The optimal solution is obtained by i When the fitness threshold is higher than the set fitness threshold and the maximum number of iterations is reached, the search is terminated and the individual global optimal position is obtained, where the individual global optimal position specifically refers to the optimal parameters of each level of the workload prediction model;

[0143] The optimal workload prediction model is obtained, specifically, the optimal workload prediction model is obtained through workload prediction model training and workload prediction model parameter optimization.

[0144] By performing the above operations, in order to solve the technical problem that the model weights and built-in parameter settings in the traditional multi-level model training optimization are inappropriate, resulting in inaccurate nursing workload prediction results, this solution uses a cross-level error compensation mechanism and an improved optimization algorithm to ensure that the errors between different levels can be effectively transmitted and compensated. The improved optimization algorithm can enhance the learning ability of each level, avoid the performance bottleneck caused by unreasonable parameter settings in traditional methods, enhance the training optimization effect of the model, and thus make the nursing workload prediction results more accurate, better meet actual needs, and significantly improve the application performance in nursing workload prediction.

[0145] Embodiment 7, see Figure 1 This embodiment is based on the above embodiment. The internal and external medicine nursing work management module specifically inputs the real-time nursing workload data of the preliminary nursing workload prediction data into the multi-dimensional workload factor refining module, refines and optimizes the workload factor through the module, obtains the optimized workload factor combination, and inputs the optimized workload factor combination into the optimal workload prediction model to obtain a multi-level nursing workload prediction result. Based on the multi-level nursing workload prediction result, hospital managers can perform accurate nursing resource allocation, scheduling optimization and nursing workload balance to improve the overall nursing management efficiency.

[0146] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0147] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0148] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A big data-based internal and external medicine nursing workload prediction system, characterized by: It includes data collection module, data preprocessing module, multi-dimensional workload factor extraction module, workload prediction model establishment module, model training optimization module and internal and external surgery nursing work management module; The data collection module obtains original data of nursing workload prediction by collecting data from the hospital management system; The data preprocessing module specifically performs data cleaning, data noise filtering, spatiotemporal data alignment and data standardization on the data to obtain preliminary data for nursing workload prediction; The multi-dimensional workload factor extraction module is used to extract multi-dimensional workload factors by constructing a workload factor pool and screening out key workload factors therefrom, extracting spatial workload factors and temporal workload factors at the same time to generate spatiotemporal workload factors, using a multi-head attention mechanism to dynamically obtain the weights of the workload factors, and merging the weights with the various workload factors to obtain an optimized workload factor combination; The workload prediction model establishment module constructs a three-level prediction architecture of a department-level macro prediction layer, a ward-level meso prediction layer, and a bed-level micro prediction layer, integrates the prediction results of the department-level macro prediction layer, the ward-level meso prediction layer, and the bed-level micro prediction layer, generates a comprehensive multi-level nursing workload prediction result, and obtains a workload prediction model; The model training optimization module optimizes the training process of the workload prediction model by introducing a cross-level error compensation mechanism, and uses an improved optimization algorithm to adjust the workload prediction model parameters to obtain an optimal workload prediction model; The internal and external medicine nursing work management module realizes nursing workload prediction based on the optimal workload prediction model and obtains multi-level nursing workload prediction results. Based on this data result, hospital managers can perform overall nursing work management and resource allocation.

2. According to the big data-based internal and external medicine nursing workload prediction system of claim 1, it is characterized by: The multi-dimensional workload factor refining module specifically includes the following steps: Construct a workload factor pool, specifically extract multidimensional workload factors from preliminary data of nursing workload prediction, and integrate them into a workload factor pool; Screening key workload factors, which is used to screen out workload factors with high impact on nursing workload prediction from the workload factor pool, specifically using the mutual information algorithm to calculate the correlation between the workload factor and the nursing workload, obtaining the workload factor with a mutual information value higher than a threshold, and obtaining the key workload factor; The generation of spatiotemporal workload factors is used to optimize the combination of spatial and temporal workload factors and capture the spatiotemporal correlation of nursing work. Specifically, the spatial workload factor and the temporal workload factor are extracted and combined to generate a new spatiotemporal workload factor, including the following steps: Spatial workload factor extraction,Based on the spatial relationship of the three-level spatial mapping model,,the spatial workload factor is generated using a graph embedding algorithm,to capture the nursing synergy between adjacent beds, Temporal workload factor extraction, using wavelet transform to extract multi-scale temporal workload factors and capture the temporal variation trend in the nursing process; To obtain the spatiotemporal workload factor, the formula used is as follows: ； In the formula, represents the jth spatiotemporal workload factor, represents the weight of the spatial workload factor, represents the weight of the time workload factor, represents the spatial workload factor, represents the time workload factor; Obtain dynamic workload factor weights so that the most relevant features can obtain higher weights in different nursing scenarios. Specifically, the weights of each key workload factor and spatiotemporal workload factor are dynamically adjusted using a multi-head attention mechanism. Generate an optimized workload factor combination, specifically by combining the dynamically adjusted workload factor weights with the key workload factor and the spatiotemporal workload factor to generate an optimized workload factor combination, and the formula used is as follows: ； In the formula, represents the optimized workload factor combination, represents the i-th critical workload factor, represents the jth spatiotemporal workload factor, represents the dynamically adjusted weight of the i-th critical workload factor, represents the dynamically adjusted weight of the jth spatiotemporal workload factor, represents the feature cross-weight parameter, Represents the Sigmoid activation function.

3. The big data-based internal and external medicine nursing workload prediction system according to claim 1 is characterized by: The module for establishing a workload prediction model is used to establish a workload prediction model to predict nursing workloads of different levels and obtain a workload prediction model; specifically, the following steps are included: Construct a department-level macro prediction layer to predict the nursing workload of the entire department. Specifically, this layer uses a bidirectional long short-term memory neural network as the prediction network of the department-level macro prediction layer to obtain the department-level workload results for the next 24 hours as the input data of the ward-level meso prediction layer. A ward-level meso-level prediction layer is constructed to predict the nursing workload of each ward. It can capture the relationship between resource usage, number of patients and nursing hours in the ward. Specifically, this layer uses a recursive neural network as the prediction network of the ward-level meso-level prediction layer to obtain the nursing workload results of the ward level in the next 12 hours as the input data of the bed-level micro-prediction layer. Construct a bed-level micro-prediction layer to predict the nursing load based on individualized bed demand and patient condition. Specifically, this layer uses a graph neural network as the prediction network of the bed-level micro-prediction layer to obtain the bed-level nursing workload results for the next 4 hours. Generate a multi-level nursing workload prediction result, specifically by integrating the prediction results of the department-level macro prediction layer, the ward-level meso prediction layer, and the bed-level micro prediction layer to generate a comprehensive multi-level nursing workload prediction result.

4. The big data-based internal and external medicine nursing workload prediction system according to claim 1 is characterized by: The model training and optimization module is used to train and optimize the workload prediction model, improve the prediction results of the model, and obtain the optimal workload prediction model; specifically, it includes the following steps: Workload prediction model training is used to ensure that the model learns and captures the patterns and trends of nursing workload. Specifically, it includes the following steps: Prepare training data, specifically use the historical nursing workload data in the preliminary data for nursing workload prediction as training data, randomly divide the training data into three equal parts, one for training the department-level macro prediction layer, another for training the ward-level meso prediction layer, and the remaining part for training the bed-level micro prediction layer; A cross-level error compensation mechanism is introduced. Specifically, in the training process of each level, the prediction error of the current level is first calculated. Through the cross-level error compensation mechanism, the error of the upstream level is passed to the downstream level, and the error of the upstream level is used as an additional constraint term of the loss function of the downstream level. The loss function is minimized through the back propagation algorithm, the weight of each level is adjusted, and the prediction effect of the model is optimized; the following steps are included: Department-level layer training, specifically, calculates the prediction error value of this layer, updates the weight of the department-level macro prediction layer through the back propagation algorithm, and transfers it to the ward-level meso prediction layer; the formula used is as follows: ； In the formula, represents the error value of the department-level macro-forecast layer, represents the total number of department-level macro prediction layer training data, represents the true value of the i-th sample, represents the predicted value of the department-level macro-forecast layer of the i-th sample; Ward-level layer training involves calculating the prediction error value of this layer, updating the weights of the ward-level meso-prediction layer through the back-propagation algorithm, and transferring the error to the bed-level micro-prediction layer. The formula used is as follows: ； In the formula, represents the error value of the ward-level meso-prediction layer, represents the total number of training data for the ward-level meso-prediction layer, represents the true value of the jth sample, represents the predicted value of the ward-level meso-level prediction layer of the jth sample, represents the weighting coefficient of the error of the macro prediction layer at the department level; Bed-level layer training, specifically, calculates the prediction error of this layer and updates the weights of the bed-level micro-prediction layer through the back-propagation algorithm; the formula used is as follows: ； In the formula, represents the error value of the bed-level micro-prediction layer, represents the total number of training data for the bed-level micro prediction layer, represents the true value of the kth sample, represents the predicted value of the bed-level micro-prediction layer of the k-th sample, represents the weighting coefficient of the ward-level meso-prediction layer error; Optimizing the parameters of the workload prediction model, specifically using an improved optimization algorithm to adjust the parameters of each level of the workload prediction model; The optimal workload prediction model is obtained, specifically, the workload prediction model is trained and the workload prediction model parameters are optimized to obtain the optimal workload prediction model.

5. The big data-based internal and external medicine nursing workload prediction system according to claim 1 is characterized by: The workload prediction model parameter optimization specifically obtains the optimal parameters of each level of the workload prediction model through a collaborative grouping optimization strategy and an optimization algorithm improved by searching for a balance factor, and includes the following steps: Initialize parameters, specifically by constructing initial algorithm parameters; The initial parameters of the algorithm include the number of individuals N and the maximum number of iterations ; Initialize the population, specifically, the individuals in the population have an initial position and initial velocity in the solution space, and the initial position and initial velocity are obtained randomly; The fitness evaluation of the individuals in the population is to calculate the fitness value F of the particles in the population. i ; The performance of each prediction layer of the workload prediction model established based on the individual position is used as the fitness value of the individual; Grouping of individuals in the population, specifically using a collaborative grouping optimization strategy to divide the population into multiple groups. Each group has a leader, whose position is the best position in the group. Individuals in each group search around the leader. Get individual search parameters using the following formula: ； In the formula, Represents the step size factor, which is used to adjust the balance between global search and local search. represents the frequency of the global best individual, Indicates the current iteration number, represents the movement frequency of the i-th individual, represents the minimum frequency of individual movement, represents the maximum frequency of individual movement, represents a random number uniformly distributed in the range [0,1], represents the search balance factor, Represents a tuning parameter that affects the decay rate of the search balance factor; Update the individual properties of the population, specifically control the search phase according to the search balance factor, the formula used is as follows: ； In the formula, represents the position of the i-th individual in the t+1-th iteration, represents the position of the i-th individual in the t-th iteration, represents the leader of the individuals in the jth group, represents the position of the rth individual in the tth iteration, represents a random variable drawn from a normal distribution, represents a random variable drawn from a standard normal distribution, represents the movement speed of the i-th individual in the t+1-th iteration, represents a random number uniformly distributed in the range [0,1], represents a random number uniformly distributed in the range [0.5,1], Represents a random number uniformly distributed in the range [-1,1], represents the search intensity of the i-th individual in the t-th iteration, represents the search capability of the i-th individual in the t-th iteration; End the current iteration, specifically by comparing the fitness values ​​to obtain the global optimal position, and replace the group with the worst fitness value in the group in the population with the average position of the remaining groups of individuals, and update the search intensity parameter and search ability parameter at the same time; the formula used is as follows: ； ； In the formula, represents the fitness function, represents the global optimal position of an individual; The optimal solution is obtained by i When it is higher than the set fitness threshold and reaches the maximum number of iterations, the search is terminated and the individual global optimal position is obtained. The individual global optimal position specifically refers to the optimal parameters of each level of the workload prediction model.

6. The big data-based internal and external medicine nursing workload prediction system according to claim 1, characterized in that: The internal and external medicine nursing work management module specifically inputs the real-time nursing workload data of the preliminary nursing workload prediction data into the multidimensional workload factor refining module, refines and optimizes the workload factor through the module, obtains the optimized workload factor combination, and inputs the optimized workload factor combination into the optimal workload prediction model to obtain multi-level nursing workload prediction results. Based on the multi-level nursing workload prediction results, hospital managers can perform accurate nursing resource allocation, scheduling optimization and nursing workload balance to improve the overall nursing management efficiency.

7. The big data-based internal and external medicine nursing workload prediction system according to claim 1 is characterized by: The data collection module specifically obtains the original data of nursing workload prediction from the hospital management system by collecting data; the original data of nursing workload prediction includes historical nursing workload data and real-time nursing workload data; the historical nursing workload data and real-time nursing workload data both include patient data, nursing staff data and department data; the historical nursing workload data also includes historical nursing workload results.

8. The big data-based internal and external medicine nursing workload prediction system according to claim 1 is characterized by: The data preprocessing module specifically performs data cleaning, data noise filtering, spatiotemporal data alignment and data standardization on the raw data of the nursing workload prediction to obtain preliminary data of the nursing workload prediction; The spatiotemporal data alignment is used to ensure the consistency of data from multiple sources in the same time frame and spatial dimension. Specifically, the dynamic time warping algorithm is used to calculate the best match of the time series, process data sources with inconsistent timestamps, ensure the time series data is time aligned, and spatially align the data from different departments, wards and beds by constructing a three-level spatial mapping model; the three-level spatial mapping model is department-ward-bed.

Images