A method, system, device and storage medium for dispatching emergency medical resources

Through the combination of deep reinforcement learning model and particle swarm algorithm, emergency medical resource scheduling is optimized, the problem of parameter setting affecting efficiency is solved, and more efficient resource scheduling strategy generation is achieved.

CN118471456BActive Publication Date: 2024-11-26GUANGDONG YITONG SOFTWARE CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410932594.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-26
Estimated Expiration
2044-07-12

AI Technical Summary

Technical Problem

In the prior art, parameter setting has a great impact on algorithm performance and convergence speed, especially when the search space is large or the structure is complex, it is difficult to determine the best parameters, resulting in low efficiency in scheduling of emergency medical resources.

Method used

Combining the deep reinforcement learning model and particle swarm algorithm, an initial scheduling scheme is constructed by optimizing the weight parameters and bias parameters of the deep reinforcement learning model, and an optimal resource scheduling strategy is obtained through iterative optimization.

Benefits of technology

Improve the performance and efficiency of resource scheduling, ensure a more comprehensive search solution space, and improve the speed and accuracy of obtaining the best parameters.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118471456B_ABST
    Figure CN118471456B_ABST
Patent Text Reader

Abstract

The present invention discloses a method, system, device and storage medium for scheduling emergency medical resources, which relates to the technical field of resource allocation, and includes the following steps: obtaining medical resource scheduling status information and related actions in the emergency room; estimating related actions through a deep reinforcement learning model; optimizing the parameters of the deep reinforcement learning model using a particle swarm algorithm, and randomly initializing the position and speed of each particle in the particle swarm; updating the speed and position of each particle according to the individual historical best position and the global historical best position; selecting the optimal particle to update the parameters of the deep reinforcement learning model to obtain an optimized deep reinforcement learning model; inputting the medical resource scheduling status information and related actions into the optimized deep reinforcement learning model to obtain the optimal resource scheduling strategy. The present invention improves the method of optimizing the training of the deep reinforcement learning model using a particle swarm algorithm, and improves the speed and accuracy of obtaining the optimal parameters in resource scheduling.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of resource allocation, and in particular to a method, system, device and storage medium for scheduling emergency medical resources. Background Art

[0002] In the process of treating critically ill patients, "time is life." Since the onset time of the patient is highly uncertain, the medical resources of the hospital emergency department, such as medical staff, equipment, space, etc., are very limited.

[0003] Under traditional conditions, people usually rely on improvisation and experience to dispatch and allocate emergency resources, which is time-consuming and labor-intensive, extremely inefficient, and easily delays rescue opportunities. Moreover, with the dynamic changes of various resource constraints, the difficulty of generating corresponding scheduling plans will increase exponentially and cannot be handled manually. Therefore, the emergency department urgently needs an intelligent scheduling method to achieve real-time output of optimized resource allocation plans.

[0004] Existing intelligent scheduling technologies are mainly based on traditional evolutionary computing methods, such as genetic algorithms and ant colony algorithms. These algorithms have relatively high computational efficiency, but are prone to fall into local optimal solutions during the search process, and the parameter settings have a large impact. Choosing appropriate parameters has a great impact on the performance and convergence speed of the algorithm, but it is usually difficult to determine the optimal parameters, especially when the search space of the problem is large or has a complex structure. Summary of the invention

[0005] The purpose of the present invention is to provide a scheduling method, system, device and storage medium for emergency medical resources in order to address the shortcomings of the above-mentioned prior art, so as to solve the problem that the parameter setting in the prior art has a great influence. The selection of appropriate parameters has a great influence on the performance and convergence speed of the algorithm, but it is usually difficult to determine the optimal parameters, especially when the search space of the problem is large or has a complex structure.

[0006] The present invention specifically provides the following technical solution: a method for dispatching emergency medical resources, comprising the following steps:

[0007] Obtain medical resource dispatch information within the emergency room;

[0008] Estimating the states and related actions in the medical resource scheduling information through a deep reinforcement learning model, and constructing an initial scheduling plan through the estimated states and related actions and reward relationships;

[0009] The weight parameters and bias parameters of the deep reinforcement learning model are optimized using a particle swarm algorithm, and the position and velocity of each particle in the particle swarm are randomly initialized, and the parameters of each particle at a certain position are used as the fitness of each particle;

[0010] Obtain the best historical position of each particle and the best global historical position according to the fitness, update the speed and position of each particle, and select the best particle according to the updated speed and position of each particle;

[0011] Update the weight parameters and bias parameters of the deep reinforcement learning model through the optimal particles to obtain the optimized deep reinforcement learning model;

[0012] The states and related actions to be scheduled are input into the optimized deep reinforcement learning model, and the optimal resource scheduling strategy is obtained by iteratively optimizing the initial scheduling plan.

[0013] Preferably, the obtaining of the medical resource dispatching information in the emergency room comprises the following steps:

[0014] Build status information of medical resource scheduling through the current number of patients in the emergency room, patient waiting time, working status of medical staff and equipment utilization rate;

[0015] Construct related actions and action spaces through actions performed under the current state of medical resource scheduling;

[0016] A reward function is generated based on reward signals generated by relevant states and relevant actions by obtaining rewards based on reduced patient waiting time, improved resource utilization, and improved patient treatment outcomes.

[0017] Preferably, the reward signal generated based on the relevant state and the relevant action generates a reward function, and the specific expression is:

[0018]

[0019] in, For reward, that is, status Next action The reward after that, R is the reward signal.

[0020] Preferably, before estimating the states and related actions in the medical resource scheduling information by using the deep reinforcement learning model, the deep reinforcement learning model needs to be trained, including the following steps:

[0021] Use historical scheduling data to collect and organize status ,action and rewards A collection of sequences ;

[0022] Initialize the deep neural network parameters θ, at the current state Next, we select an action based on the current strategy, by A greedy strategy is used to balance exploration and utilization; the deep neural network parameters θ include weight parameters and bias parameters;

[0023] Execute an action , observe the reward and the new state ;

[0024] Get the Q value of the target network , and use the mean square error MSE as the loss function to update the deep neural network parameters θ to obtain the minimized loss function.

[0025] Preferably, the Q value of the target network is obtained , and use the mean square error MSE as the loss function to update the deep neural network parameter θ to obtain the minimized loss function, including the following steps:

[0026] Get the Q value of the target network , the specific expression is:

[0027]

[0028] Where γ is the discount factor, θ′ is the parameter of the target network, is the action taken in the new state;

[0029] Using mean square error MSE as the loss function, the specific expression is:

[0030]

[0031] Update the deep neural network parameters θ to obtain the minimized loss function, the specific expression is:

[0032]

[0033] Where α is the learning rate, N is the number of samples in the batch, is the target Q value when the number of samples is i, To find the partial derivative of the loss function L(θ);

[0034] Update the deep neural network parameters θ′ of the target network to the values ​​of the deep neural network parameters θ of the current network.

[0035] Preferably, the parameter of each particle at a certain position is used as the fitness of each particle, and the specific expression is:

[0036]

[0037] in, , v id is the velocity of particle i in the dth dimension, v id+1is the velocity of particle i in the dth dimension next time, x id is the position of particle i in the dth dimension, x id+1 is the next position of particle i in the dth dimension, ω is the inertia weight, c1 and c2 are different learning factors, pbest id is the best historical position of particle i, gbest id is the best historical position of the entire population, and rand1 and rand2 are both random numbers in [0, 1].

[0038] Preferably, after the optimal resource scheduling strategy is obtained by iteratively optimizing the initial scheduling scheme, the hyperparameters of the deep reinforcement learning model are adjusted, and the hyperparameters include a learning rate and a discount factor.

[0039] The present invention also provides a dispatching system for emergency medical resources, comprising:

[0040] A collection module is used to obtain the medical resource dispatch information in the emergency room;

[0041] A model building module, used to estimate the states and related actions in the medical resource scheduling information through a deep reinforcement learning model, and to build an initial scheduling plan through the estimated states and related actions and reward relationships;

[0042] A particle acquisition module, used to optimize the parameters of the deep reinforcement learning model using a particle swarm algorithm, and randomly initialize the position and velocity of each particle in the particle swarm, and use the parameters of each particle at a certain position as the fitness of each particle;

[0043] An updating module, used for obtaining the best historical position of each particle and the best global historical position according to the fitness, updating the speed and position of each particle, and selecting the best particle according to the updated speed and position of each particle;

[0044] An optimization module is used to update the weight parameters and bias parameters of the deep reinforcement learning model through the optimal particles to obtain an optimized deep reinforcement learning model;

[0045] The optimal scheduling module is used to input the states and related actions to be scheduled into the optimized deep reinforcement learning model, and obtain the optimal resource scheduling strategy by iteratively optimizing the initial scheduling plan.

[0046] The present invention also provides a computer device, including a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of the method for scheduling emergency medical resources.

[0047] The present invention also provides a storage medium on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of the method for scheduling emergency medical resources are implemented.

[0048] Compared with the prior art, the present invention has the following significant advantages:

[0049] The present invention obtains the status information and related actions of medical resource scheduling, constructs a deep reinforcement learning model through a deep neural network, and optimizes the parameters of the deep reinforcement learning model through a particle swarm algorithm, that is, combining deep reinforcement learning technology with the particle swarm algorithm, and proposes a method for improving and optimizing the deep reinforcement learning model using the particle swarm algorithm, thereby improving the training effect of the deep reinforcement learning model. After training, the medical resource scheduling status information in the emergency room is used as the input of the deep reinforcement model, so that better performance and effect can be obtained when solving the resource scheduling problem in the field of emergency medicine. The particle swarm algorithm can ensure the solving efficiency of the deep reinforcement learning model, can search the solution space more comprehensively, and improve the speed and accuracy of obtaining the best parameters in resource scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0051] The following is a clear and complete description of the technical solutions of the embodiments of the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0052] Deep reinforcement learning (DRL) is a new technology that combines deep learning and reinforcement learning. It learns the optimal strategy through the interaction between the agent and the environment. In intelligent scheduling, deep reinforcement learning is often used to solve problems such as dynamic scheduling and real-time decision-making. This method will open up new ideas, adapt deep reinforcement learning to the scheduling of emergency medical resources, and further combine a traditional evolutionary computing method (particle swarm optimization algorithm) to optimize its scheduling model and scheduling strategy, fully integrating the advantages of traditional methods and deep reinforcement learning, thereby improving the accuracy and efficiency of scheduling plan generation.

[0053] like Figure 1 As shown, an embodiment of the present invention provides a method for scheduling emergency medical resources, comprising the following steps:

[0054] Step S1: Obtain the medical resource scheduling information in the emergency room.

[0055] Specifically:

[0056] The medical resource scheduling status information is constructed through the current number of patients in the emergency room, patient waiting time, medical staff's working status and equipment utilization.

[0057] The state vector can be a vector containing this information. In the emergency medical resource scheduling problem, the state space can represent the current resource utilization of the hospital, including but not limited to:

[0058] The working status of medical staff: including busy, idle, performing tasks, etc.

[0059] Usage status of medical equipment: including occupied status, idle status, etc.

[0060] Information such as the patient's waiting time and the severity of their condition.

[0061] The current state is , ,in Represents the state space.

[0062] Construct related actions and action space through the actions performed in the current state.

[0063] The action space can represent the actions that can be performed in the current state, for example:

[0064] Assign healthcare workers to specific patients.

[0065] Arrange the patient for further examination or treatment.

[0066] Adjust the use of medical equipment, etc.

[0067] The set of possible actions is , indicating that in the state The following set of optional actions.

[0068] By obtaining rewards based on reduced patient waiting time, improved resource utilization, and improved patient treatment outcomes, a reward function is generated based on the reward signals generated by relevant states and relevant actions, reflecting the quality of the decision.

[0069] The reward function is expressed as follows:

[0070]

[0071] in, For reward, that is, status Next action After the reward.

[0072] The reward function is used to evaluate how well the agent performs a specific action in a specific state. In the emergency medical resource scheduling problem, the reward function can be defined based on the following indicators:

[0073] Reduce patient waiting time.

[0074] Improve the utilization rate of medical resources.

[0075] Improve patient treatment outcomes, etc.

[0076] Step S2: Estimate the states and related actions in the medical resource scheduling information through a deep reinforcement learning model, and construct an initial scheduling plan through the estimated states and related actions and reward relationships.

[0077] Specifically:

[0078] Deep reinforcement learning (DRL), as an important branch of artificial intelligence, combines deep learning and reinforcement learning techniques and is widely used in various complex decision-making problems. In the medical field, especially in challenging scenarios such as emergency medical resource scheduling, deep reinforcement learning technology has great potential and can help hospitals improve resource utilization efficiency, shorten patient waiting time, and optimize medical service quality. This paper will explore in depth how to use deep reinforcement learning technology to achieve intelligent scheduling of emergency medical resources, including problem modeling, model design, data preparation, training and optimization, real-time decision-making, and model evaluation.

[0079] Deep reinforcement learning (DRL) can handle problems with high-dimensional state space and action space. Specifically, in this invention, for the emergency medical resource scheduling problem, a deep Q-network is used to approximate the state value function or action value function, thereby achieving the learning and execution of the optimal strategy. Using a neural network to estimate the action value function , the deep neural network parameter θ represents the weight of the neural network.

[0080] After using a deep neural network to build a deep reinforcement learning model, the deep reinforcement learning model is trained, including the following steps:

[0081] Use historical data to collect and organize the sequence of states, actions, and rewards .

[0082] Initialize the deep neural network parameters θ, at the current state Next, we select an action based on the current strategy, by A greedy strategy is used to balance exploration and exploitation.

[0083] Execute an action , observe the reward and the new state .

[0084] Get the target Q value , and use the mean square error MSE as the loss function to update the deep neural network parameters θ to obtain the minimized loss function.

[0085]

[0086] Where γ is the discount factor, θ′ is the parameter of the target network, is the action taken in the new state.

[0087] Using mean square error MSE as the loss function, the specific expression is:

[0088]

[0089] Update the deep neural network parameters θ to minimize the loss function. The specific expression is:

[0090]

[0091] Where α is the learning rate, N is the number of samples in the batch, is the target Q value when the number of samples is i, To find the partial derivative of the loss function L(θ).

[0092] Update the deep neural network parameters θ′ of the target network to the values ​​of the deep neural network parameters θ of the current network.

[0093] Step S3: Use the particle swarm algorithm to optimize the weight parameters and bias parameters of the deep reinforcement learning model, and randomly initialize the position and velocity of each particle in the particle swarm, and use the parameters of each particle at a certain position as the fitness of each particle.

[0094] Particle swarm optimization (PSO) is an optimization algorithm based on swarm intelligence that simulates the foraging behavior of bird flocks. Its core idea is to maintain a group of particles to search for the optimal solution in the solution space. Each particle updates its position and speed based on its own historical experience and the group's optimal experience to find the global optimal solution.

[0095] The core update formula of the PSO algorithm, that is, the specific expression of the particle's fitness is:

[0096]

[0097] in, , v id is the velocity of particle i in the dth dimension, v id+1 is the velocity of particle i in the dth dimension next time, x id is the position of particle i in the dth dimension, x id+1 is the next position of particle i in the dth dimension, ω is the inertia weight, c1 and c2 are different learning factors, pbest id is the best historical position of particle i, gbest id is the best historical position of the entire population, and rand1 and rand2 are both random numbers in [0, 1].

[0098] Step S4: Obtain the best historical position of each particle and the global best historical position according to the fitness, update the speed and position of each particle, and select the optimal particle according to the updated speed and position of each particle.

[0099] Step S5: Update the weight parameters and bias parameters of the deep reinforcement learning model through the optimal particles to obtain an optimized deep reinforcement learning model.

[0100] Step S6: Input the states to be scheduled and related actions into the optimized deep reinforcement learning model, and obtain the optimal resource scheduling strategy by iteratively optimizing the initial scheduling plan.

[0101] After obtaining the optimal resource scheduling strategy through iterative optimization, adjust the hyperparameters of the DRL model, including the learning rate and discount factor, to further improve the performance of the model.

[0102] Based on the above method, the present invention provides a scheduling system for emergency medical resources, including a collection module, a model building module, a particle acquisition module, an update module, an optimization module and an optimal scheduling module.

[0103] Among them, the acquisition module is used to obtain the medical resource scheduling information in the emergency room; the model construction module is used to use a deep neural network to build a deep reinforcement learning model, and estimate the state and related actions in the medical resource scheduling information through the deep reinforcement learning model to obtain the initial scheduling plan; the particle acquisition module is used to use the particle swarm algorithm to optimize the parameters of the deep reinforcement learning model, and randomly initialize the position and speed of each particle in the particle swarm, and take the parameters of each particle at a certain position as the fitness of each particle; the update module is used to obtain the historical best position and the global historical best position of each particle according to the fitness, update the speed and position of each particle, and select the optimal particle through the updated speed and position of each particle; the optimization module is used to update the weight parameters and bias parameters of the deep reinforcement learning model through the optimal particles to obtain the optimized deep reinforcement learning model; the optimal scheduling module is used to input the scheduling state information and related actions of the medical resources to be scheduled into the optimized deep reinforcement learning model, and obtain the optimal resource scheduling strategy by iteratively optimizing the initial scheduling plan.

[0104] The present invention also provides a computer device, including a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a method for scheduling emergency medical resources.

[0105] According to the disclosed embodiments, a computing device may communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth communications, etc.), or with any device (e.g., routers, modems, etc.) that enables a computing device to communicate with one or more other computing devices.

[0106] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, a method for scheduling emergency medical resources is implemented.

[0107] According to the disclosed embodiments, the storage medium may be a non-volatile computer-readable storage medium, such as but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0108] The above content is a further detailed description of the present invention in combination with a specific preferred embodiment. For technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for dispatching emergency medical resources. It is characterized in that The steps include: Obtain medical resource dispatch information within the emergency room; Estimating the states and related actions in the medical resource scheduling information through a deep reinforcement learning model, and constructing an initial scheduling plan through the estimated states and related actions and reward relationships; The weight parameters and bias parameters of the deep reinforcement learning model are optimized using a particle swarm algorithm, and the position and velocity of each particle in the particle swarm are randomly initialized, and the parameters of each particle at a certain position are used as the fitness of each particle; Obtain the best historical position of each particle and the best global historical position according to the fitness, update the speed and position of each particle, and select the best particle according to the updated speed and position of each particle; Update the weight parameters and bias parameters of the deep reinforcement learning model through the optimal particles to obtain the optimized deep reinforcement learning model; Inputting the states and related actions to be scheduled into the optimized deep reinforcement learning model, and obtaining the optimal resource scheduling strategy by iteratively optimizing the initial scheduling scheme; The step of obtaining the medical resource dispatching information in the emergency room includes the following steps: Build status information of medical resource scheduling through the current number of patients in the emergency room, patient waiting time, working status of medical staff and equipment utilization rate; Construct related actions and action spaces through actions performed under the current state of medical resource scheduling; Generate a reward function based on reward signals generated by relevant states and relevant actions by obtaining rewards based on reduced patient waiting time, improved resource utilization, and improved patient treatment outcomes; Before estimating the states and related actions in the medical resource scheduling information by using the deep reinforcement learning model, the deep reinforcement learning model needs to be trained, including the following steps: Use the historical data of the schedule to collect and organize the status t 、Action a t and reward r t The sequence set (s t ,a t ,r t ); Initialize the deep neural network parameters θ, at the current state s t Next, an action is selected according to the current strategy; the deep neural network parameters θ include weight parameters and bias parameters; Execute action a t , observe the reward r t and the new state s t+1 ; Get the Q value y of the target network t , and use the mean square error MSE as the loss function to update the deep neural network parameters θ to obtain the minimized loss function.

2. A method for dispatching emergency medical resources as claimed in claim 1, It is characterized in that The reward signal generated based on the relevant state and the relevant action generates a reward function, and the specific expression is: r t =R(s t ,a t ) Among them, r t is the reward, i.e., state s t Next, perform action a t The reward after that, R is the reward signal.

3. A method for dispatching emergency medical resources as claimed in claim 1, It is characterized in that The Q value y of the target network is obtained t , and use the mean square error MSE as the loss function to update the deep neural network parameter θ to obtain the minimized loss function, including the following steps: Get the Q value y of the target network t , the specific expression is: Where γ is the discount factor, θ′ is the parameter of the target network, and a′ is the action taken in the new state; Using mean square error MSE as the loss function, the specific expression is: Update the deep neural network parameters θ to obtain the minimized loss function, the specific expression is: Among them, α is the learning rate, N is the number of samples in the batch, and y i is the target Q value when the number of samples is i, To find the partial derivative of the loss function L(θ); Update the deep neural network parameters θ′ of the target network to the values ​​of the deep neural network parameters θ of the current network.

4. A method for dispatching emergency medical resources as claimed in claim 1, It is characterized in that The parameter of each particle at a certain position is used as the fitness of each particle. The specific expression is: v id+1 =ω·v id +c 1 ·edge 1 ·(pbest id -x id )+c 2 ·edge 2 ·(gbest id -x id ) Among them, x id+1 =x id +v id+1 , v id is the velocity of particle i in the dth dimension, v id+1 is the velocity of particle i in the dth dimension next time, x id is the position of particle i in the dth dimension, x id+1 is the next position of particle i in the dth dimension, ω is the inertia weight, c 1 and c 2 are different learning factors, pbest id is the best historical position of particle i, gbest id is the best historical position of the entire population, rand 1 and rand 2 are all random numbers in [0,1].

5. A method for dispatching emergency medical resources as claimed in claim 1, It is characterized in that After the optimal resource scheduling strategy is obtained by iteratively optimizing the initial scheduling plan, the hyperparameters of the deep reinforcement learning model are adjusted, and the hyperparameters include a learning rate and a discount factor.

6. A dispatching system for emergency medical resources, It is characterized in that include A collection module is used to obtain the medical resource dispatch information in the emergency room; A model building module, used to estimate the states and related actions in the medical resource scheduling information through a deep reinforcement learning model, and to build an initial scheduling plan through the estimated states and related actions and reward relationships; A particle acquisition module, used to optimize the parameters of the deep reinforcement learning model using a particle swarm algorithm, and randomly initialize the position and velocity of each particle in the particle swarm, and use the parameters of each particle at a certain position as the fitness of each particle; An updating module, used for obtaining the best historical position of each particle and the best global historical position according to the fitness, updating the speed and position of each particle, and selecting the best particle according to the updated speed and position of each particle; An optimization module is used to update the weight parameters and bias parameters of the deep reinforcement learning model through the optimal particles to obtain an optimized deep reinforcement learning model; An optimal scheduling module, used to input the states and related actions to be scheduled into the optimized deep reinforcement learning model, and obtain the optimal resource scheduling strategy by iteratively optimizing the initial scheduling plan; The collection module obtains the medical resource dispatch information in the emergency room, including the following steps: Build status information of medical resource scheduling through the current number of patients in the emergency room, patient waiting time, working status of medical staff and equipment utilization rate; Construct related actions and action spaces through actions performed under the current state of medical resource scheduling; Generate a reward function based on reward signals generated by relevant states and relevant actions by obtaining rewards based on reduced patient waiting time, improved resource utilization, and improved patient treatment outcomes; Before the model building module estimates the states and related actions in the medical resource scheduling information through the deep reinforcement learning model, the deep reinforcement learning model needs to be trained, including the following steps: Use historical scheduling data to collect and organize status t 、Action a t and reward r t The sequence set (s t ,a t ,r t ); Initialize the deep neural network parameters θ, at the current state s t Next, an action is selected according to the current strategy; the deep neural network parameters θ include weight parameters and bias parameters; Execute action a t , observe the reward r t and the new state s t+1 ; Get the Q value y of the target network t , and use the mean square error MSE as the loss function to update the deep neural network parameters θ to obtain the minimized loss function.

7. A computer device, It is characterized in that It includes a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a method for scheduling emergency medical resources as described in any one of claims 1 to 5.

8. A storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of a method for scheduling emergency medical resources according to any one of claims 1 to 5 are implemented.

Citation Information

Cited By

  • Medical resource allocation method and device for pre-hospital first aid, electronic equipment and storage medium

    CN120299653A