Hospital operation efficiency management method and system based on CMP value

Abstract

The invention discloses a hospital operation efficiency management method and system based on a CMP value, and belongs to the technical field of hospital management, and the method comprises the steps of hospital operation data collection, hospital operation data preprocessing, hospital operation efficiency evaluation model construction, model parameter optimization and hospital operation efficiency management. According to the scheme, the optimal number of neighbor nodes is dynamically determined, a topological structure and feature information are fused to construct a node similarity matrix, enhanced features are obtained by aggregating the neighbor nodes, node features are learned through a double-layer feature learning network, and similarity comparison loss is introduced to set a total loss function; according to an iterative search progress proportion and population diversity, a dynamic adjustment coefficient, a weight coefficient and a disturbance coefficient are designed, an elite layer is updated based on the three coefficients, an exploration layer is updated based on an all-body optimal position and a random individual position, a restart mechanism is set, a CMP value is introduced, and the accuracy of hospital operation efficiency evaluation is improved. And reliable support is provided for management decision.

Description

Technical Field

[0001] The present invention belongs to the field of hospital management technology, and specifically refers to a hospital operation efficiency management method and system based on CMP value. Background Art

[0002] The hospital operation efficiency management method is a method that uses deep learning and big data analysis technology to integrate the hospital's multi-dimensional information to evaluate and optimize the overall operation efficiency of medical institutions, thereby improving medical service efficiency, reducing operating costs, and improving patients' medical experience, ultimately achieving the goal of efficient use of medical resources and refined hospital management.

[0003] However, the existing hospital operation efficiency management methods are unable to dynamically capture the similarities and nonlinear complex relationships between the operation data of different hospitals. A single scale feature is difficult to fully reflect the operation rules of the hospital, resulting in inaccurate hospital operation efficiency evaluation results and deviations in management strategies. The existing hospital operation efficiency management methods are insufficiently sensitive to hospital operation data and are difficult to adapt to the dynamic changes in hospital operation data, resulting in distorted evaluation results and ineffective management decisions. The existing hospital operation efficiency management methods have the problem that the CMI value as the mean is easily affected by extreme values, while the total CM value focuses on the overall amount, resulting in inaccurate hospital operation efficiency evaluation. Summary of the Invention

[0004] In view of the above situation, in order to overcome the defects of the existing technology, the present invention provides a hospital operation efficiency management method and system based on CMP value, which is aimed at the problem that the existing hospital operation efficiency management method cannot dynamically capture the similarity and nonlinear complex relationship between different hospital operation data, and the single scale feature is difficult to fully reflect the hospital operation law, resulting in inaccurate hospital operation efficiency evaluation results and management strategy deviation. This solution dynamically determines the optimal number of neighbor nodes based on node weight and label consistency, constructs an adaptive neighborhood graph, and can accurately reflect the nonlinear complex relationship between data; extracts multi-scale features, fuses topological structure and feature information to construct a node similarity matrix, which can more comprehensively measure the similarity between hospital operation data; obtains enhanced features by aggregating neighbor nodes, enriches its own feature expression; learns node features through a two-layer feature learning network, and can effectively extract deep-level information in node features; introduces similarity comparison loss to set the total loss function, improves the accuracy of hospital operation efficiency evaluation, enhances the adaptability of hospital management strategy and reduces hospital management cost; The efficiency management method is insufficiently sensitive to hospital operating data, making it difficult to adapt to the dynamic changes in hospital operating data, resulting in distorted evaluation results and ineffective management decisions. This solution designs dynamic adjustment coefficients, weight coefficients, and perturbation coefficients based on the iterative search progress ratio and population diversity, which can quickly adapt to the dynamic changes in hospital operating data. The population is divided into an elite layer and an exploration layer, and the elite layer is updated based on the three coefficients, while the exploration layer is updated based on the optimal position of the entire group and the position of random individuals, which can continuously optimize the evaluation model parameters. The fitness change rate is calculated, a restart mechanism is set to find the optimal parameters, and the model's adaptation to the data is monitored in real time to ensure the accuracy of the evaluation results and provide reliable support for management decisions. In order to address the problem in existing hospital operating efficiency management methods that the CMI value, as the mean, is easily affected by extreme values, while the CM total value focuses on the overall quantity, resulting in inaccurate hospital operating efficiency evaluation, this solution introduces the CMP value, which takes into account both workload and efficiency. This can more comprehensively and objectively evaluate the hospital's operating conditions, staffing, and other conditions, providing a scientific basis for hospital management.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a hospital operation efficiency management method based on CMP value, which includes the following steps:

[0006] Step S1: hospital operation data collection;

[0007] Step S2: preprocessing of hospital operation data;

[0008] Step S3: Constructing a hospital operation efficiency evaluation model;

[0009] Step S4: model parameter optimization;

[0010] Step S5: Hospital operation efficiency management.

[0011] Furthermore, in step S1, the hospital operation data collection is to collect historical hospital operation data; the historical hospital operation data includes CMP value, resource input data, medical quality data, operation efficiency data, patient experience data and efficiency level;

[0012] The CMP value refers to the weighted value created per capita in the hospital, CMP value = total CM value / total number of relevant medical staff; the total CM value is the sum of the weights of all cases in the hospital within a month; the total number of relevant medical staff is the number of medical staff involved in medical services in the hospital during that period.

[0013] Furthermore, in step S2, the hospital operation data preprocessing is to perform data cleaning, data normalization, data encoding and data set construction on the data to obtain a training data set and a test data set.

[0014] Furthermore, in step S3, the hospital operation efficiency evaluation model is constructed based on a deep neural network, specifically including the following steps:

[0015] Step S31: Construct an adaptive neighborhood graph; take each data in the training data set as a node, and the node feature is the corresponding hospital operation data, and construct a node feature matrix; pre-set the value range of the number of neighbor nodes Q is , traverse the Q value and calculate the neighbor quality of each node under each Q value; for each node, select the Q value that maximizes the neighbor quality as the optimal number of neighbor nodes Q of the node best ; According to the optimal number of neighbor nodes, each node is connected to its corresponding neighbor nodes with edges to construct an adaptive neighborhood graph and adjacency matrix; where N is the number of data in the training data set, is rounded down; calculating the neighbor quality includes the following steps:

[0016] Step S311: Select neighbor nodes; calculate the Euclidean distance between the feature vectors of any two nodes, perform the fourth root operation on the Euclidean distance, then process it through the hyperbolic tangent function, and finally subtract 1 from the distance. , get the weight between two nodes; for each node p i , sort all other nodes in descending order according to their weights, and select the first Q nodes as nodes p i Neighbor nodes; among them, D ij is node p i and node p j The Euclidean distance between i and p j are the i-th and j-th nodes respectively, i and j are node indexes, Q is the number of neighboring nodes, is the hyperbolic tangent function;

[0017] Step S312: Calculate neighbor quality; for each node p i , check whether the labels of its Q neighbor nodes are the same as those of node p i If they are the same, the weight of the neighboring node is accumulated, otherwise the weight of the neighboring node is ignored; calculate the node p i The average weight between its Q neighbor nodes, the product of the adjustment coefficient and the average weight is used as the neighbor penalty coefficient β, and the sum of the weights of neighbors with the same label is divided by Q to the power of β to obtain the neighbor quality of each node;

[0018] Step S32: constructing a node similarity matrix; calculating a feature similarity matrix and a topology similarity matrix according to the node feature matrix and the adjacency matrix, respectively, and combining the feature similarity matrix and the topology similarity matrix to obtain a node similarity matrix; comprising the following steps:

[0019] Step S321: Calculate the feature similarity matrix; apply 3×3 and 7×7 convolution kernels to extract the local detail feature matrix X from the node feature matrix local and the global trend feature matrix X global , calculate X respectively local and X global The variance of X is obtained by normalization local and X global The importance weight of X local and X global Weighted fusion obtains the multi-scale feature matrix X fused ; Calculate the cosine similarity value of the multi-scale features between any two nodes, construct the cosine similarity matrix, normalize each row of the cosine similarity matrix, and obtain the feature similarity matrix ;

[0020] Step S322: Calculate the topological similarity matrix; normalize each row of the adjacency matrix to obtain the topological similarity matrix ;

[0021] Step S33: Enhance features; for each node p i , weighted aggregation of the features of its neighboring nodes to obtain aggregated features, and then combine the aggregated features with the features of its own nodes to obtain enhanced features and construct the enhanced feature matrix E;

[0022] Step S34: a two-layer feature learning network; the two-layer feature learning network adopts a deep neural network architecture, including two hidden layers and one output layer, the first hidden layer is composed of a fully connected layer FC, a ReLU activation function and a Dropout regularization, the second hidden layer is composed of a fully connected layer and a ReLU activation function, and the output layer is composed of a fully connected layer and a Softmax activation function; the enhanced feature matrix is ​​input into the first hidden layer to obtain a preliminary learning feature matrix, the preliminary learning feature matrix is ​​then input into the second hidden layer to obtain a secondary learning feature matrix, the secondary learning feature matrix is ​​input into the output layer to obtain an evaluation result of each node;

[0023] Step S35: Set the total loss function; for each node p i , the feature similarity index of all node pairs is weighted as the numerator through the node similarity matrix, the sum of the feature similarity index of all non-self nodes is used as the denominator, and the negative logarithm is taken to obtain the similarity contrast loss; the average similarity contrast loss of all nodes is used as the similarity contrast loss of the model; then the cross entropy loss of the model is calculated, and the similarity contrast loss and cross entropy loss are combined through the balance coefficient to obtain the total loss of the model.

[0024] Furthermore, in step S4, the model parameter optimization specifically includes the following steps:

[0025] Step S41: initial individual positions; establishing a parameter search space for the learnable parameters in the hospital operation efficiency evaluation model, where the learnable parameters are the weight matrix and bias term of the two-layer feature learning network in the hospital operation efficiency evaluation model. Randomly initialize A individual positions in the parameter search space, and use the A individual positions as a population. Each individual position represents a set of learnable parameters, and the total loss of the hospital operation efficiency evaluation model established based on the learnable parameters on the test data set is used as the fitness value of the corresponding individual position.

[0026] Step S42: setting three coefficients; obtaining a dynamic adjustment coefficient based on the iterative search progress ratio and population diversity; combining the dynamic adjustment coefficient with a sine function to obtain a dynamic weight coefficient; combining the dynamic adjustment coefficient with a sine and cosine function to obtain a perturbation convergence factor; combining the perturbation convergence factor with the upper and lower limits of the parameter search space to obtain a random perturbation coefficient;

[0027] Step S43: Hierarchical co-evolution update; all individual positions are arranged in ascending order of fitness value, the population is stratified, the first 30% of individual positions are used as the elite layer, and the remaining individual positions are used as the exploration layer; the individual positions in the elite layer are updated based on the three set coefficients; the individual positions in the exploration layer are updated based on the overall optimal position and the random individual position;

[0028] Step S44: Set a restart mechanism; pre-set the fitness threshold and change rate threshold. When the fitness value of an individual position is less than the fitness threshold, the learnable parameter represented by the individual position is used as the optimal parameter, and a hospital operation efficiency evaluation model is established based on the optimal parameter; otherwise, after 0.5T iterative searches, calculate the fitness change rate of 0.1T consecutive iterative searches. If the fitness change rate is less than the change rate threshold or reaches the maximum number of iterative searches, restart and return to step S41 to re-initialize the individual position; otherwise, the number of iterative searches is increased by 1 and return to step S42 to continue the iterative search.

[0029] Furthermore, in step S5, the hospital operation efficiency management is to collect real-time hospital operation data, and the real-time hospital operation data includes CMP value, resource input data, medical quality data, operational efficiency data and patient experience data; after pre-processing the real-time hospital operation data, it is input into the hospital operation efficiency evaluation model established based on the optimal parameters for evaluation, and according to the output evaluation results, the efficiency level of the real-time hospital operation data is obtained, and the corresponding management strategy is generated to manage the hospital.

[0030] The present invention provides a hospital operation efficiency management system based on CMP value, which includes a hospital operation data acquisition module, a hospital operation data preprocessing module, a hospital operation efficiency evaluation model construction module, a model parameter optimization module and a hospital operation efficiency management module;

[0031] The hospital operation data acquisition module collects historical hospital operation data and sends the data to the hospital operation data preprocessing module;

[0032] The hospital operation data preprocessing module performs data cleaning, data normalization, data encoding and data set construction to obtain a training data set and a test data set, and sends the data to the hospital operation efficiency evaluation model construction module;

[0033] The hospital operation efficiency evaluation model module dynamically determines the optimal number of neighbor nodes based on node weights and label consistency, constructs an adaptive neighborhood graph, extracts multi-scale features, fuses topological structure and feature information to construct a node similarity matrix, obtains enhanced features by aggregating neighbor nodes, learns node features through a two-layer feature learning network, outputs evaluation results, introduces similarity comparison loss to set the total loss function, and sends the data to the model parameter optimization module;

[0034] The model parameter optimization module designs dynamic adjustment coefficients, weight coefficients, and perturbation coefficients based on the iterative search progress ratio and population diversity, divides the population into an elite layer and an exploration layer, updates the elite layer based on the three coefficients, updates the exploration layer based on the overall optimal position and random individual positions, calculates the fitness change rate, sets a restart mechanism, finds the optimal parameters, and sends the data to the hospital operation efficiency management module;

[0035] The hospital operation efficiency management module collects and pre-processes real-time hospital operation data, obtains the efficiency level and generates corresponding management strategies based on the evaluation results output by the hospital operation efficiency evaluation model established based on optimal parameters.

[0036] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0037] (1) In view of the problem that the existing hospital operation efficiency management methods cannot dynamically capture the similarity and nonlinear complex relationship between the operation data of different hospitals, and the single-scale feature is difficult to fully reflect the hospital operation rules, resulting in inaccurate hospital operation efficiency evaluation results and management strategy deviations, this scheme dynamically determines the optimal number of neighbor nodes based on node weight and label consistency, constructs an adaptive neighborhood graph, so that the data in the neighborhood have higher similarity and correlation, and can accurately reflect the nonlinear complex relationship between the data; extracts multi-scale features, so that the model can more comprehensively and deeply describe the hospital operation rules; integrates topological structure and feature information to construct a node similarity matrix, which can more comprehensively measure the similarity between hospital operation data; enhances features by aggregating neighbor nodes, so that each node can obtain the operation information of similar nodes and enrich its own feature expression; learns node features through a two-layer feature learning network and outputs evaluation results. Through multi-layer nonlinear transformation, it can effectively extract deep-level information in node features; introduces similarity comparison loss to set the total loss function, improves the accuracy of hospital operation efficiency evaluation, enhances the adaptability of hospital management strategies and reduces hospital management costs.

[0038] (2) In view of the problem that the existing hospital operation efficiency management methods are not sensitive enough to hospital operation data and are difficult to adapt to the dynamic changes of hospital operation data, resulting in distorted evaluation results and ineffective management decisions, this scheme randomly generates multiple individual positions, takes the total loss as the fitness value, comprehensively covers different parameter combinations, and directly relates the model's fitting effect on actual operation data, thereby improving sensitivity to data changes; according to the iterative search progress ratio and population diversity, dynamic adjustment coefficients, weight coefficients and perturbation coefficients are designed to enhance the global search capability in the early stage of change, focus on local optimization in the later stage, and introduce randomness, which can quickly adapt to the dynamic changes of hospital operation data; the population is divided into an elite layer and an exploration layer, the elite layer is updated based on the three coefficients, the exploration layer is updated based on the optimal position of the whole group and the random individual position, the elite layer is fine-tuned, and the exploration layer is extensively searched, which can continuously optimize the evaluation model parameters; the fitness change rate is calculated, a restart mechanism is set, the optimal parameters are found, and the model's adaptation to the data is monitored in real time to ensure the accuracy of the evaluation results, provide reliable support for management decisions, and ensure that management decisions are always based on accurate evaluation results.

[0039] (3) In view of the problem in the existing hospital operation efficiency management method that the CMI value as the mean is easily affected by extreme values, and the total CM value focuses on the overall quantity, resulting in inaccurate evaluation of hospital operation efficiency, this plan introduces the CMP value, which takes into account both workload and efficiency. By introducing the CMP value, it can more comprehensively and objectively evaluate the hospital's operation status, personnel allocation, etc., and provide a scientific basis for hospital management. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of a process for managing hospital operation efficiency based on CMP value provided by the present invention;

[0041] Figure 2 A schematic diagram of a hospital operation efficiency management system based on CMP value provided by the present invention;

[0042] Figure 3 Schematic diagram of the process of step S3;

[0043] Figure 4 Schematic diagram of the process of step S4.

[0044] 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

[0045] 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.

[0046] In the description of the present invention, it should be understood 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. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0047] Example 1, see Figure 1 The present invention provides a hospital operation efficiency management method based on CMP value, which includes the following steps:

[0048] Step S1: Hospital operation data collection: collect historical hospital operation data;

[0049] Step S2: preprocessing of hospital operation data; data cleaning, data normalization, data encoding and data set construction to obtain training data set and test data set;

[0050] Step S3: Construct a hospital operation efficiency evaluation model; dynamically determine the optimal number of neighbor nodes based on node weight and label consistency, construct an adaptive neighborhood graph, extract multi-scale features, fuse topological structure and feature information to construct a node similarity matrix, obtain enhanced features by aggregating neighbor nodes, learn node features through a two-layer feature learning network, output evaluation results, and introduce similarity comparison loss to set the total loss function;

[0051] Step S4: Model parameter optimization: Based on the iterative search progress ratio and population diversity, design dynamic adjustment coefficients, weight coefficients, and perturbation coefficients, divide the population into an elite layer and an exploration layer, update the elite layer based on the three coefficients, and update the exploration layer based on the overall optimal position and random individual positions. Calculate the fitness change rate, set a restart mechanism, and find the optimal parameters.

[0052] Step S5: Hospital operation efficiency management; collect real-time hospital operation data and pre-process it, obtain the efficiency level and generate the corresponding management strategy based on the evaluation results output by the hospital operation efficiency evaluation model established based on the optimal parameters.

[0053] Example 2, see Figure 1This embodiment is based on the above embodiment. In step S1, the hospital operation data collection is to collect historical hospital operation data; the historical hospital operation data includes CMP value, resource input data, medical quality data, operation efficiency data, patient experience data and efficiency level;

[0054] The resource input data include the number of beds, equipment value and average working hours per person;

[0055] The medical quality data include cure rates, complication rates, and readmission rates;

[0056] Said operational efficiency data include average length of stay and operating room turnover rate;

[0057] Said patient experience data includes satisfaction scores, complaint rates, and average wait times for appointments;

[0058] The efficiency levels include high efficiency, medium efficiency and low efficiency;

[0059] The CMP value refers to the weighted value created per capita by the hospital, CMP value = total CM value / total number of related medical staff; total CM value is the sum of the weights of all cases in a hospital within a certain period; the total number of related medical staff is the number of medical staff involved in medical services in the hospital during that period; this scheme calculates the CMP value based on the total CM value and the total number of related medical staff within a month;

[0060] DRG is a medical service management tool that can provide relatively scientific and effective evaluation indicators from three aspects: medical service capacity, service efficiency, and quality and safety. The medical service capacity evaluation indicators based on DRG include the number of DRG groups, the case combination index CMI value, and the total weight CM value; the CMI value can reflect the complexity of the cases admitted to the hospital, but as a mean it is easily affected by extreme values, and the CM total value is used to measure the overall complexity of the patients admitted to the hospital and the consumption of medical resources. It focuses on the overall volume but cannot effectively evaluate the operating efficiency; the CMP value combines the advantages of the CMI value and the CM total value, avoiding the distortion problem of the CMI value as the mean, and the disadvantages of the CM total value that can only evaluate the volume but not the efficiency. It takes into account both workload and efficiency. By introducing the CMP value, it can more comprehensively and objectively evaluate the hospital's operation, personnel allocation, etc., and provide a scientific basis for hospital management.

[0061] Example 3, see Figure 1 ,This embodiment is based on the above embodiment.,In step S2, the hospital operation data preprocessing is to perform data cleaning, data normalization, data encoding and data set construction on the data to obtain a training data set and a test data set;

[0062] The data cleaning is to remove the error values, missing values ​​and abnormal values ​​in the data;

[0063] The data normalization is to unify the numerical data into the same range using the maximum-minimum scaling method;

[0064] The data encoding is to convert categorical data into numerical data using One-Hot encoding;

[0065] The data set is constructed by using the efficiency level as a data label, and then constructing a training data set and a test data set based on the data after data cleaning, data normalization and data encoding processing.

[0066] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S3, the hospital operation efficiency evaluation model is constructed based on a deep neural network, specifically including the following contents:

[0067] Step S31: Construct an adaptive neighborhood graph; hospital operation data is multidimensional and has large distribution differences. Traditional fixed neighborhood methods are difficult to dynamically capture the similarity differences between data, resulting in a lack of pertinence in the selection of neighborhood nodes. Adaptive neighborhood can accurately characterize the characteristic associations of similar hospitals, and adaptively select the optimal number of neighbors for each node. By calculating the neighbor quality, the weights of neighbors with the same label are preferentially retained, and irrelevant nodes are filtered to make the neighborhood structure more stable; each data in the training data set is regarded as a node, and the node feature is the corresponding hospital operation data, and a node feature matrix is ​​constructed; the value range of the number of neighbor nodes Q is pre-set to be , traverse the Q value and calculate the neighbor quality of each node under each Q value; for each node, select the Q value that maximizes the neighbor quality as the optimal number of neighbor nodes Q of the node best ; According to the optimal number of neighbor nodes, each node is connected to its corresponding neighbor node with edges to construct an adaptive neighborhood graph and adjacency matrix; where N is the number of data in the training data set, is rounded down; calculating the neighbor quality includes the following steps:

[0068] Step S311: Select neighbor nodes; the dimensions of hospital operation data are complex, and directly using Euclidean distance may not reflect business relevance, resulting in distortion in neighbor node selection. Through the fourth root operation and hyperbolic tangent function, the absolute difference of distance in high-dimensional space is weakened, and the similarity of low-difference data is strengthened. Neighbors are selected in descending order of weight to ensure that the nodes in the neighborhood are most similar in comprehensive operation characteristics; calculate the Euclidean distance between the feature vectors of any two nodes, perform the fourth root operation on the Euclidean distance, and then process it through the hyperbolic tangent function, and finally subtract 1 from it. , get the weight between two nodes; for each node p i, sort all other nodes in descending order according to their weights, and select the first Q nodes as nodes p i Neighbor nodes; among them, D ij is node p i and node p j The Euclidean distance between i and p j are the i-th and j-th nodes respectively, i and j are node indexes, Q is the number of neighboring nodes, is the hyperbolic tangent function;

[0069] Step S312: Calculate the quality of neighbors. In the evaluation of hospital operation efficiency, labels are the core guide, but traditional methods may introduce cross-label interference nodes, resulting in model deviation. By using the indicator function to only accumulate the weights of neighbors with the same label, it is ensured that the nodes in the neighborhood have the same efficiency level attributes and avoid heterogeneous interference. The neighbor penalty coefficient is combined with the average weight. When the overall similarity of the neighborhood is low, the penalty intensity of Q is increased to suppress the influence of low-quality neighbors. For each node p i , check whether the labels of its Q neighbor nodes are the same as those of node p i If they are the same, the weight of the neighboring node is accumulated, otherwise the weight of the neighboring node is ignored; calculate the node p i The average weight between its Q neighbor nodes is calculated, and the product of the adjustment coefficient and the average weight is used as the neighbor penalty coefficient β. The sum of the weights of neighbors with the same label is divided by Q raised to the power of β to obtain the neighbor quality of each node. The formula used is as follows:

[0070] ;

[0071] Where Z i is node p i Neighbor quality, ω iq is node p i and node p q The weight between q is node p i The qth neighbor node of , q is the neighbor node index, y i and y q They are node p i and node p q Data labels; is the indicator function, if the data labels are the same, the value is 1, otherwise it is 0; β is the neighbor penalty coefficient, , α is the adjustment coefficient in the range of (0.5, 1), is node p i The average weight of its Q neighbor nodes;

[0072] Step S32: Construct a node similarity matrix; hospital operation data contains both local details and global trends. A single scale feature is difficult to fully reflect the operation rules, and traditional methods ignore the impact of topological structure on similarity. Local and global features are extracted separately, and after weighted fusion, multi-level operation modes are captured. Combining the adjacency matrix and the feature matrix, both the direct relationship between data and the intrinsic similarity of data features are considered to improve the comprehensiveness of the similarity measurement; the feature similarity matrix is ​​calculated based on the node feature matrix and the adjacency matrix. and the topological similarity matrix , combining the feature similarity matrix and the topology similarity matrix to obtain the node similarity matrix ; including the following steps:

[0073] Step S321: Calculate the feature similarity matrix; apply 3×3 and 7×7 convolution kernels to extract the local detail feature matrix X from the node feature matrix local and the global trend feature matrix X global , calculate X respectively local and X global The variance of X is obtained by normalization local and X global The importance weight of X local and X global Weighted fusion obtains the multi-scale feature matrix X fused ; Calculate the cosine similarity value of the multi-scale features between any two nodes, construct the cosine similarity matrix, normalize each row of the cosine similarity matrix, and obtain the feature similarity matrix ;

[0074] Step S322: Calculate the topological similarity matrix; normalize each row of the adjacency matrix to obtain the topological similarity matrix ;

[0075] Step S33: Enhance features; the features of a single node may lack contextual information. Traditional methods ignore the neighborhood synergy effect. By weighted summing and aggregating neighbor features, each node contains the operational information of its similar data. Combining aggregated features with its own features, it avoids excessive reliance on the neighborhood and causes the loss of individual characteristics, and balances commonality and individuality. For each node p i , weightedly aggregate the features of its neighboring nodes to obtain aggregated features, and then combine the aggregated features with the features of its own node to obtain enhanced features and construct the enhanced feature matrix E; the formula used is as follows:

[0076] ;

[0077] Where h i is node p i The aggregation characteristics of is node p i The optimal number of neighbor nodes, ω iq is node p i and node p q The weight between q is node p q characteristics;

[0078] Step S34: Two-layer feature learning network; hospital operation efficiency evaluation needs to deal with nonlinear complex relationships. Traditional linear models are difficult to capture deep laws and have a high risk of overfitting. The first hidden layer extracts nonlinear features through ReLU and Dropout, and suppresses overfitting through random inactivation. The second hidden layer further refines high-order features. The output layer maps the features into efficiency level probability distribution to provide richer evaluation results; the two-layer feature learning network adopts a deep neural network architecture, including two hidden layers and one output layer. The first hidden layer is composed of a fully connected layer FC, a ReLU activation function and a Dropout regularization. The second hidden layer is composed of a fully connected layer and a ReLU activation function. The output layer is composed of a fully connected layer and a Softmax activation function. The enhanced feature matrix is ​​input into the first hidden layer to obtain a preliminary learning feature matrix, and then the preliminary learning feature matrix is ​​input into the second hidden layer to obtain a secondary learning feature matrix. The secondary learning feature matrix is ​​input into the output layer to obtain the evaluation result of each node. The formula used is as follows:

[0079] ;

[0080] ;

[0081] ;

[0082] Among them, R1 and R2 are the initial learning feature matrix and the secondary learning feature matrix respectively. is the evaluation result;

[0083] Step S35: Set the total loss function; the traditional evaluation model only focuses on classification accuracy and ignores the similarity constraints between samples, resulting in poor generalization ability of the model, especially poor performance in data imbalance scenarios. The contrast loss is used to force the secondary learning features of similar data to be more similar and heterogeneous data to be further apart. The ratio of classification loss to contrast loss is adjusted by μ. When the data is balanced, the classification accuracy is emphasized. When there are few samples in the minority class, the similarity constraint is strengthened to avoid the model being biased towards the majority class. For each node p i, the feature similarity index of all node pairs is weighted by the node similarity matrix as the numerator, the sum of the feature similarity indexes of all non-self nodes is used as the denominator, and the negative logarithm is taken to obtain the similarity contrast loss; the average similarity contrast loss of all nodes is used as the similarity contrast loss of the model; then the cross entropy loss of the model is calculated, and the similarity contrast loss and cross entropy loss are combined through the balance coefficient to obtain the total loss of the model; the formula used is as follows:

[0084] ;

[0085] ;

[0086] in, is node p i Similarity comparison loss, 、 and They are node p i , node p j and node p v The corresponding features in the secondary learning feature matrix are, is the cosine similarity function, is node p i and node p j Node similarity, Loss CE and Loss ass are the cross entropy loss and total loss of the model, is the temperature coefficient in the range of (0.5, 1.5), and μ is the balance coefficient in the range of [0.1, 0.5].

[0087] By performing the above operations, we address the problems in existing hospital operation efficiency management methods, such as the inability to dynamically capture the similarities and nonlinear complex relationships between different hospital operation data, and the difficulty of single-scale features in fully reflecting hospital operation patterns, which leads to inaccurate hospital operation efficiency evaluation results and management strategy deviations. This solution dynamically determines the optimal number of neighbor nodes based on node weights and label consistency, constructs an adaptive neighborhood graph, and ensures that the data within the neighborhood have higher similarity and correlation, accurately reflecting the nonlinear complex relationships between data. Multi-scale features are extracted, making the model more comprehensive and in-depth in characterizing hospital operation patterns. A node similarity matrix is ​​constructed by fusing topological structure and feature information, which can more comprehensively measure the similarity between hospital operation data and make the evaluation results more consistent with the actual operation of the hospital. Enhanced features are obtained by aggregating neighboring nodes, allowing each node to obtain the operation information of similar nodes and enrich its own feature expression. Node features are learned through a two-layer feature learning network and output evaluation results. Multi-layer nonlinear transformations can effectively extract deep information from node features. Similarity comparison loss is introduced to set the total loss function, improving the accuracy of hospital operation efficiency evaluation, enhancing the adaptability of hospital management strategies, and reducing hospital management costs.

[0088] Example 5, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S4, the model parameter optimization specifically includes the following steps:

[0089] Step S41: initial individual position; the parameter space of the hospital operation efficiency evaluation model is complex and nonlinear, and the traditional gradient descent method is prone to fall into local optimality and cannot find the global optimal solution, especially in high-dimensional and heterogeneous scenarios of medical data. It is difficult to optimize. By randomly initializing A individual positions, searching simultaneously at multiple points in the parameter space, the impact of different parameter configurations on model performance is captured, and the total loss of the test set is used as the fitness value, which is directly related to the performance of the model in the actual evaluation, ensuring that the optimization goal is consistent with the hospital management needs; establishing a parameter search space for the learnable parameters in the hospital operation efficiency evaluation model, the learnable parameters are the weight matrix and bias term of the two-layer feature learning network in the hospital operation efficiency evaluation model, and randomly initializing A individual positions in the parameter search space. The A individual positions are used as a population, and each individual position represents a set of learnable parameters. The total loss of the hospital operation efficiency evaluation model established based on the learnable parameters on the test data set is used as the fitness value of the corresponding individual position;

[0090] Step S42: Set three coefficients; the fixed parameters of the traditional optimization algorithm are difficult to adapt to the dynamic changes of hospital data, resulting in slow convergence speed in the later stage of the search or jumping out of the optimal solution. The dynamic adjustment coefficient dynamically scales the search step according to the iteration progress and population diversity, automatically adapts to the high-dimensional characteristics of medical data, avoids falling into the local optimum, and improves the search efficiency of the complex parameter space. The dynamic weight coefficient modulates the adjustment coefficient through the sine function, periodically changes the parameter update intensity, and enhances the ability to jump out of the local extreme value. The random perturbation coefficient introduces controllable randomness in the parameter space to prevent the algorithm from converging prematurely, especially for the nonlinear relationship in medical data, and suppresses data noise interference; the dynamic adjustment coefficient is obtained according to the iterative search progress ratio and population diversity; the dynamic adjustment coefficient is combined with the sine function to obtain the dynamic weight coefficient; the dynamic adjustment coefficient is combined with the sine and cosine functions to obtain the perturbation convergence factor; the perturbation convergence factor is combined with the upper and lower limits of the parameter search space to obtain the random perturbation coefficient; the formula used is as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] ;

[0095] ;

[0096] Where, J t 、M t , K t 、H t and G t They are the population diversity, dynamic adjustment coefficient, dynamic weight coefficient, perturbation convergence factor and random perturbation coefficient at the t-th iterative search, T is the maximum number of iterative searches, t is the iterative search index, and are the sine and cosine functions, is a random number generation function, UB and LB are the upper and lower limits of the parameter search space, C is the dimension of the individual position, is the fitness value of the ath individual position during the tth iteration search, where a is the individual position index;

[0097] Step S43: Hierarchical co-evolution update; a single optimization strategy is difficult to take into account both global search and local optimization at the same time, especially in the high-dimensional and sparse scenario of hospital data, which is prone to premature convergence or oscillatory non-convergence problems. The first 30% of individuals are optimized separately, and the local optimum is further explored through dynamic weights and random perturbations, which is suitable for steadily improving the parameter configuration that has performed well. The remaining individuals are guided by the overall optimal and random elites to search for potential solutions in a wider space; all individual positions are arranged in ascending order according to the size of the fitness value, and the population is stratified. The first 30% of individual positions are used as the elite layer, and the remaining individual positions are used as the exploration layer; the individual positions in the elite layer are updated based on the three set coefficients; the individual positions in the exploration layer are updated based on the overall optimal position and the random individual position; the formula used is as follows:

[0098] ;

[0099] ;

[0100] Where, and are the a1th individual position in the elite layer during the t+1th and tth iteration searches, respectively. is the a2th individual position in the exploration layer during the t+1th iteration search, a1 and a2 are individual position indices, is the optimal position of the whole group during the t-th iterative search. The optimal position of the whole group is the individual position with the smallest fitness value. and are two different individual positions randomly selected from the elite layer during the t-th iteration search, 、 and They are 、 and The fitness value of and Otherwise and The square of is the smoothing term, ;

[0101] Step S44: Set a restart mechanism; the optimization process may stagnate due to falling into a local optimum or a change in data distribution. Traditional algorithms lack self-correction capabilities and require manual intervention to restart. When the fitness change rate is lower than the threshold, it indicates that the algorithm has converged or stagnated. At this time, restarting can introduce a new random initial point to jump out of the local optimum; pre-set the fitness threshold and change rate threshold. When the fitness value of an individual position is less than the fitness threshold, the learnable parameter represented by the individual position is used as the optimal parameter, and a hospital operation efficiency evaluation model is established based on the optimal parameter; otherwise, after 0.5T iterative searches, calculate the fitness change rate of 0.1T consecutive iterative searches. If the fitness change rate is less than the change rate threshold or reaches the maximum number of iterative searches, restart and return to step S41 to re-initialize the individual position; otherwise, the number of iterative searches is increased by 1 and return to step S42 to continue the iterative search; the formula used is as follows:

[0102] ;

[0103] Where, is the fitness change rate during the t-th iteration search, It is the fitness value of the optimal position of the whole in the t-0.1Tth iterative search.

[0104] By performing the above operations, the existing hospital operation efficiency management methods are insufficiently sensitive to hospital operation data, making it difficult to adapt to the dynamic changes of hospital operation data, resulting in distorted evaluation results and ineffective management decisions. This solution randomly generates multiple individual positions, uses total loss as the fitness value, comprehensively covers different parameter combinations, and directly relates the model's fit to the actual operation data, thereby improving sensitivity to data changes. Based on the iterative search progress ratio and population diversity, dynamic adjustment coefficients, weight coefficients, and perturbation coefficients are designed to enhance global search capabilities in the early stages of changes, focus on local optimization in the later stages, and introduce randomness to quickly adapt to dynamic changes in hospital operation data. The population is divided into an elite layer and an exploration layer, and the elite layer is updated based on three coefficients. The exploration layer is updated based on the optimal position of the entire group and random individual positions. The elite layer undergoes fine-tuning to adapt to gradual changes in the data, while the exploration layer conducts extensive searches to promptly capture new patterns brought about by data mutations, thereby continuously optimizing the evaluation model parameters. The fitness change rate is calculated, a restart mechanism is set, the optimal parameters are found, and the model's adaptation to the data is monitored in real time to ensure the accuracy of the evaluation results, provide reliable support for management decisions, and ensure that management decisions are always based on accurate evaluation results.

[0105] Example 6, see Figure 1This embodiment is based on the above embodiment. In step S5, hospital operation efficiency management collects real-time hospital operation data, which includes CMP value, resource input data, medical quality data, operational efficiency data and patient experience data. After preprocessing the real-time hospital operation data, it is input into a hospital operation efficiency evaluation model established based on optimal parameters for evaluation. According to the output evaluation results, the efficiency level of the real-time hospital operation data is obtained, and the corresponding management strategy is generated to manage the hospital.

[0106] By performing the above operations, in order to address the problem in existing hospital operation efficiency management methods that the CMI value as the mean is easily affected by extreme values, while the total CM value focuses on the overall volume, resulting in inaccurate evaluation of hospital operation efficiency, this solution introduces the CMP value, which takes both workload and efficiency into consideration. By introducing the CMP value, it is possible to more comprehensively and objectively evaluate the hospital's operation status, personnel allocation, etc., providing a scientific basis for hospital management.

[0107] Example 7, see Figure 2 This embodiment is based on the above embodiment. The present invention provides a hospital operation efficiency management system based on CMP value, including a hospital operation data acquisition module, a hospital operation data preprocessing module, a hospital operation efficiency evaluation model construction module, a model parameter optimization module and a hospital operation efficiency management module;

[0108] The hospital operation data acquisition module collects historical hospital operation data and sends the data to the hospital operation data preprocessing module;

[0109] The hospital operation data preprocessing module performs data cleaning, data normalization, data encoding and data set construction to obtain a training data set and a test data set, and sends the data to the hospital operation efficiency evaluation model construction module;

[0110] The hospital operation efficiency evaluation model module dynamically determines the optimal number of neighbor nodes based on node weights and label consistency, constructs an adaptive neighborhood graph, extracts multi-scale features, fuses topological structure and feature information to construct a node similarity matrix, obtains enhanced features by aggregating neighbor nodes, learns node features through a two-layer feature learning network, outputs evaluation results, introduces similarity comparison loss to set the total loss function, and sends the data to the model parameter optimization module;

[0111] The model parameter optimization module designs dynamic adjustment coefficients, weight coefficients, and perturbation coefficients based on the iterative search progress ratio and population diversity, divides the population into an elite layer and an exploration layer, updates the elite layer based on the three coefficients, updates the exploration layer based on the overall optimal position and random individual positions, calculates the fitness change rate, sets a restart mechanism, finds the optimal parameters, and sends the data to the hospital operation efficiency management module;

[0112] The hospital operation efficiency management module collects and pre-processes real-time hospital operation data, obtains the efficiency level and generates corresponding management strategies based on the evaluation results output by the hospital operation efficiency evaluation model established based on optimal parameters.

[0113] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0114] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0115] The present invention and its embodiments are described above. This 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 a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A hospital operation efficiency management method based on CMP value, characterized by: The method comprises the following steps: Step S1: Hospital operation data collection: collect historical hospital operation data; Step S2: preprocessing of hospital operation data; data cleaning, data normalization, data encoding and data set construction to obtain training data set and test data set; Step S3: Construct a hospital operation efficiency evaluation model; dynamically determine the optimal number of neighbor nodes based on node weight and label consistency, construct an adaptive neighborhood graph, extract multi-scale features, fuse topological structure and feature information to construct a node similarity matrix, obtain enhanced features by aggregating neighbor nodes, learn node features through a two-layer feature learning network, output evaluation results, and introduce similarity comparison loss to set the total loss function; Step S4: Model parameter optimization: Based on the iterative search progress ratio and population diversity, design dynamic adjustment coefficients, weight coefficients, and perturbation coefficients, divide the population into an elite layer and an exploration layer, update the elite layer based on the three coefficients, and update the exploration layer based on the overall optimal position and random individual positions. Calculate the fitness change rate, set a restart mechanism, and find the optimal parameters. Step S5: Hospital operation efficiency management; collect real-time hospital operation data and pre-process it, obtain the efficiency level and generate the corresponding management strategy based on the evaluation results output by the hospital operation efficiency evaluation model established based on the optimal parameters.

2. The hospital operation efficiency management method based on CMP value according to claim 1, characterized in that: In step S3, the hospital operation efficiency evaluation model is constructed based on a deep neural network, specifically including the following steps: Step S31: constructing an adaptive neighborhood graph; Step S32: constructing a node similarity matrix; calculating a feature similarity matrix and a topology similarity matrix according to the node feature matrix and the adjacency matrix, respectively, and combining the feature similarity matrix and the topology similarity matrix to obtain a node similarity matrix; comprising the following steps: Step S321: Calculate the feature similarity matrix; apply 3×3 and 7×7 convolution kernels to extract the local detail feature matrix X from the node feature matrix local and the global trend feature matrix X global , calculate X respectively local and X global The variance of X is obtained by normalization local and X global The importance weight of X local and X global Weighted fusion obtains the multi-scale feature matrix X fused ; Calculate the cosine similarity value of the multi-scale features between any two nodes, construct the cosine similarity matrix, normalize each row of the cosine similarity matrix, and obtain the feature similarity matrix ; Step S322: Calculate the topological similarity matrix; normalize each row of the adjacency matrix to obtain the topological similarity matrix ; Step S33: Enhance features; for each node p i , weighted aggregation of the features of its neighboring nodes to obtain aggregated features, and then combine the aggregated features with the features of its own nodes to obtain enhanced features and construct the enhanced feature matrix E; Step S34: a two-layer feature learning network; the two-layer feature learning network adopts a deep neural network architecture, including two hidden layers and one output layer; the enhanced feature matrix is ​​input into the first hidden layer to obtain a preliminary learning feature matrix, the preliminary learning feature matrix is ​​then input into the second hidden layer to obtain a secondary learning feature matrix, and the secondary learning feature matrix is ​​input into the output layer to obtain the evaluation result of each node; Step S35: Set the total loss function.

3. The hospital operation efficiency management method based on CMP value according to claim 2, characterized in that: In step S31, the adaptive neighborhood graph is constructed; each data in the training data set is regarded as a node, and the node feature is the corresponding hospital operation data, and a node feature matrix is ​​constructed; The range of the number of neighbor nodes Q is pre-set to , traverse the Q value and calculate the neighbor quality of each node under each Q value; for each node, select the Q value that maximizes the neighbor quality as the optimal number of neighbor nodes Q of the node best ; According to the optimal number of neighbor nodes, each node is connected to its corresponding neighbor node with edges to construct an adaptive neighborhood graph and adjacency matrix; where N is the number of data in the training data set, is rounded down; calculating the neighbor quality includes the following steps: Step S311: Select neighbor nodes; calculate the Euclidean distance between the feature vectors of any two nodes, perform the fourth root operation on the Euclidean distance, then process it through the hyperbolic tangent function, and finally subtract 1 from the distance. , get the weight between two nodes; for each node p i , sort all other nodes in descending order according to their weights, and select the first Q nodes as nodes p i Neighbor nodes; among them, D ij is node p i and node p j The Euclidean distance between i and p j are the i-th and j-th nodes respectively, i and j are node indexes, Q is the number of neighboring nodes, is the hyperbolic tangent function; Step S312: Calculate neighbor quality; for each node p i , check whether the labels of its Q neighbor nodes are the same as those of node p i If they are the same, the weight of the neighboring node is accumulated, otherwise the weight of the neighboring node is ignored; calculate the node p i The average weight between the nodes and their Q neighboring nodes is used, and the product of the adjustment coefficient and the average weight is used as the neighbor penalty coefficient β. The sum of the weights of the neighbors with the same label is divided by Q to the power of β to obtain the neighbor quality of each node.

4. The hospital operation efficiency management method based on CMP value according to claim 2, characterized in that: In step S35, the total loss function is set; for each node p i , the feature similarity index of all node pairs is weighted as the numerator through the node similarity matrix, the sum of the feature similarity index of all non-self nodes is used as the denominator, and the negative logarithm is taken to obtain the similarity contrast loss; the average similarity contrast loss of all nodes is used as the similarity contrast loss of the model; then the cross entropy loss of the model is calculated, and the similarity contrast loss and cross entropy loss are combined through the balance coefficient to obtain the total loss of the model.

5. The hospital operation efficiency management method based on CMP value according to claim 1, characterized in that: In step S4, the model parameter optimization specifically includes the following steps: Step S41: initial individual positions; establishing a parameter search space for the learnable parameters in the hospital operation efficiency evaluation model, where the learnable parameters are the weight matrix and bias term of the two-layer feature learning network in the hospital operation efficiency evaluation model. Randomly initialize A individual positions in the parameter search space, and use the A individual positions as a population. Each individual position represents a set of learnable parameters, and the total loss of the hospital operation efficiency evaluation model established based on the learnable parameters on the test data set is used as the fitness value of the corresponding individual position. Step S42: setting three coefficients; obtaining a dynamic adjustment coefficient based on the iterative search progress ratio and population diversity; combining the dynamic adjustment coefficient with a sine function to obtain a dynamic weight coefficient; combining the dynamic adjustment coefficient with a sine and cosine function to obtain a perturbation convergence factor; combining the perturbation convergence factor with the upper and lower limits of the parameter search space to obtain a random perturbation coefficient; Step S43: Hierarchical co-evolution update; all individual positions are arranged in ascending order of fitness value, the population is stratified, the first 30% of individual positions are used as the elite layer, and the remaining individual positions are used as the exploration layer; the individual positions in the elite layer are updated based on the three set coefficients; the individual positions in the exploration layer are updated based on the overall optimal position and the random individual position; Step S44: Set a restart mechanism; pre-set the fitness threshold and the change rate threshold. When the fitness value of an individual position is less than the fitness threshold, the learnable parameter represented by the individual position is used as the optimal parameter, and a hospital operation efficiency evaluation model is established based on the optimal parameter; otherwise, after 0.5T iterative searches, calculate the fitness change rate of 0.1T consecutive iterative searches. If the fitness change rate is less than the change rate threshold or reaches the maximum number of iterative searches, restart and return to step S41 to re-initialize the individual position; otherwise, the number of iterative searches is increased by 1 and returns to step S42 to continue the iterative search; where T is the maximum number of iterative searches.

6. The hospital operation efficiency management method based on CMP value according to claim 1, characterized in that: In step S1, the hospital operation data collection is to collect historical hospital operation data; the historical hospital operation data includes CMP value, resource input data, medical quality data, operation efficiency data, patient experience data and efficiency level; The CMP value refers to the weighted value created per capita in the hospital, CMP value = total CM value / total number of relevant medical staff; the total CM value is the sum of the weights of all cases in the hospital within a month; the total number of relevant medical staff is the number of medical staff involved in medical services in the hospital during that period.

7. The hospital operation efficiency management method based on CMP value according to claim 1, characterized in that: In step S5, the hospital operation efficiency management is to collect real-time hospital operation data, and the real-time hospital operation data includes CMP value, resource input data, medical quality data, operation efficiency data and patient experience data; After preprocessing, the real-time hospital operation data is input into the hospital operation efficiency evaluation model established based on the optimal parameters for evaluation. According to the output evaluation results, the efficiency level of the real-time hospital operation data is obtained, and the corresponding management strategy is generated to manage the hospital.

8. The hospital operation efficiency management method based on CMP value according to claim 1, characterized in that: In step S2, the hospital operation data preprocessing is to perform data cleaning, data normalization, data encoding and data set construction to obtain a training data set and a test data set.

9. A hospital operation efficiency management system based on CMP value, used to implement a hospital operation efficiency management method based on CMP value according to any one of claims 1 to 8, characterized in that: It includes hospital operation data collection module, hospital operation data preprocessing module, hospital operation efficiency evaluation model building module, model parameter optimization module and hospital operation efficiency management module; The hospital operation data acquisition module collects historical hospital operation data and sends the data to the hospital operation data preprocessing module; The hospital operation data preprocessing module performs data cleaning, data normalization, data encoding and data set construction to obtain a training data set and a test data set, and sends the data to the hospital operation efficiency evaluation model construction module; The hospital operation efficiency evaluation model module dynamically determines the optimal number of neighbor nodes based on node weights and label consistency, constructs an adaptive neighborhood graph, extracts multi-scale features, fuses topological structure and feature information to construct a node similarity matrix, obtains enhanced features by aggregating neighbor nodes, learns node features through a two-layer feature learning network, outputs evaluation results, introduces similarity comparison loss to set the total loss function, and sends the data to the model parameter optimization module; The model parameter optimization module designs dynamic adjustment coefficients, weight coefficients, and perturbation coefficients based on the iterative search progress ratio and population diversity, divides the population into an elite layer and an exploration layer, updates the elite layer based on the three coefficients, updates the exploration layer based on the overall optimal position and random individual positions, calculates the fitness change rate, sets a restart mechanism, finds the optimal parameters, and sends the data to the hospital operation efficiency management module; The hospital operation efficiency management module collects and pre-processes real-time hospital operation data, obtains the efficiency level and generates corresponding management strategies based on the evaluation results output by the hospital operation efficiency evaluation model established based on optimal parameters.

Images