A high-value medical consumable spare part demand forecasting method

Abstract

The application discloses a high-value medical consumable spare part demand prediction method, and belongs to the technical field of high-value medical consumable spare part demand prediction; the method comprises the following steps: firstly, historical consumption data of the high-value medical consumables are acquired, and normalization processing is performed on the historical consumption data; secondly, the improved red fox optimization algorithm is used to optimize the weight and threshold of a BP neural network prediction model, and a high-value medical consumable spare part demand prediction model is constructed; finally, the historical consumption data are used as a training set to train the prediction model, the training result is subjected to inverse normalization, and a prediction result is output. The application adds reflection learning and elite reverse strategies in the red fox optimization algorithm, so that the improved algorithm has improved convergence speed and accuracy; the red fox optimization algorithm is used to optimize the BP neural network parameters, the influence of the non-ideal prediction result and the long learning time caused by the randomization of the standard BP neural network parameters is reduced, and the application can provide accurate prediction services for high-value medical consumable spare part demand prediction.

Description

Technical Field

[0001] This invention belongs to the field of high-value medical consumables demand forecasting technology, specifically relating to a method for forecasting the demand for high-value medical consumables. Background Technology

[0002] High-value medical consumables are essential for ensuring the normal operation of medical work. Forecasting spare parts demand is fundamental for medical institutions in managing spare parts procurement, supply, storage, and transportation. A reasonable and scientific spare parts inventory can effectively help decision-makers take appropriate management measures for high-value medical consumables, achieving the goals of reducing inventory costs and ensuring timely availability of spare parts.

[0003] Demand forecasting for high-value medical consumables and spare parts involves predicting future consumption based on past consumption patterns and current inventory levels. Commonly used forecasting methods include time series models, grey models, and support vector machine regression models. Because demand forecasting is a non-linear dynamic problem, and the factors influencing consumable consumption are complex, model building is relatively complicated. Among these methods, time series models are suitable for linear forecasting, grey models are suitable for predicting monotonically increasing and decreasing smooth data samples, and support vector machine regression performs poorly with large samples, requiring data preprocessing and parameter tuning. Its forecasting results for high-value medical consumables vary significantly, and its accuracy is unsatisfactory.

[0004] Currently, the demand for high-value medical consumables is quite complex, and medical institutions mostly rely on rough manual calculations to predict the demand for these consumables. These calculations are often inaccurate and fail to provide a scientific and effective way to optimize the inventory structure of high-value medical consumables spare parts, leading to increased inventory management costs and even a waste of medical resources. Therefore, there is an urgent need for an effective forecasting method to be applied to the inventory management of high-value medical consumables spare parts.

[0005] This invention provides a method for predicting the demand for high-value medical consumables, based on an improved Red Fox optimization algorithm and a BP neural network, which features high prediction accuracy and strong learning and generalization capabilities. Summary of the Invention

[0006] The purpose of this invention is to provide a method for predicting the demand for high-value medical consumables, so as to solve the problems mentioned in the background art, such as the low accuracy of manual calculation and prediction by intelligent models.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for predicting the demand for high-value medical consumables spare parts, the method comprising the following steps:

[0009] S1: Obtain high-value medical consumable consumption data over a period of time, analyze the factors affecting spare parts, and normalize the data to map the original data to the [0,1] interval.

[0010] S2: Constructing a demand prediction model for high-value medical consumables spare parts based on BP neural network: Determine the network topology based on the number of input and output parameters, determine the number of network training iterations, training objectives, learning rate, and select the BP neural network transfer function and training method.

[0011] S3: Initialize the weights and threshold of the BP neural network, setting them empirically as follows: .

[0012] S4: Obtain the optimal weights and thresholds of the BP neural network based on the improved Red Fox optimization algorithm.

[0013] S5: Construct a new demand prediction model for high-value medical consumables spare parts based on the optimal weights and thresholds.

[0014] S6: Use the historical data of high-value medical consumables and spare parts consumption within a certain time period as the training set to train and test the high-value medical consumables and spare parts demand prediction model. After inverse normalization of the prediction results, output the prediction results.

[0015] Preferably, step S2, which involves constructing a demand prediction model for high-value medical consumables spare parts based on a BP neural network, specifically includes:

[0016] S21: Determine the number of network layers. Theoretically, it has been proven that a single-hidden-layer BP neural network can achieve arbitrary nonlinear mappings by appropriately increasing the number of neurons. In most cases, a single hidden layer is sufficient. This invention uses a single hidden layer.

[0017] S22: Analyze and determine the number of factors affecting the demand for high-value medical consumables spare parts, determine the number of neurons in the input layer and output layer of the BP neural network, and combine with S21 to determine the topology of the BP neural network.

[0018] The number of neurons in the hidden layer is determined empirically.

[0019]

[0020] in, This represents the number of neurons in the hidden layer. and These represent the number of neurons in the input layer and the output layer, respectively. yes The constant.

[0021] S23: Selection of transfer function and training method; hidden layer adopts... The output layer uses a linear function, and the commonly used gradient descent method is chosen for training.

[0022] S24: Determine the number of network training iterations, training objectives, and learning rate.

[0023] S25: Obtain training sample data and train the network.

[0024] Preferably, step S4, which involves obtaining the optimal weights and thresholds of the BP neural network based on the improved Red Fox optimization algorithm, specifically includes:

[0025] S41: Initialize the red fox population, and use the error between the spare parts prediction result and the actual value as the fitness function of the red fox algorithm.

[0026] S42: Select the position of the red fox individual with the best fitness value, simulate the red fox's life habits such as foraging, hunting, and escaping hunters according to the improved red fox optimization algorithm, and iteratively update the position of the red fox individual through its five stages of finding prey, reflecting and learning, attacking prey, elite reverse strategy, and escaping hunters and developing the population. Select the best individual after iteration until the maximum number of iterations is reached or the fitness requirement is met.

[0027] S43: If the termination condition is not met, eliminate the worst individual in the red fox population and use the best individual to reproduce offspring. Repeat step S42 until the iteration ends, and obtain the best individual position as the optimal weight and threshold of the BP neural network.

[0028] Preferably, the improved Red Fox optimization algorithm in step S42 specifically includes:

[0029] S421: Algorithm Assumptions: The individual positions of the red fox population are initialized as follows: ,in The number of individuals in the population. , , , Let be the dimension of the individual's position; Indicates the first During the nth iteration The location of the red fox is determined. A fitness function is pre-defined, assuming the red foxes move within a specific spatial range. The red fox with the best fitness is recorded as the optimal individual, referred to as the "optimal individual". fox.

[0030] S422: Calculate the fitness value and select the red fox individual with the lowest fitness. The mean square error (MSE) between the predicted and actual values ​​in the BP neural network training set is chosen as the fitness function for the red fox individual.

[0031]

[0032] in This represents the actual demand for high-value medical consumables. To predict the demand for high-value medical consumables, the smaller the fitness value of a red fox individual, the higher the prediction accuracy of the algorithm.

[0033] S423: Hunting Prey – Global Search Phase. In this phase, individual red foxes venture out to hunt prey. The fox pack communicates with each other, sharing the easiest locations to catch prey, causing the pack to move towards the optimal individual's location. If the fitness value after moving is better than the original location, the location is updated; otherwise, it returns to the original location. In one iteration, the individual and the optimal individual... The distance can be expressed as:

[0034]

[0035] The formula for updating individual location is:

[0036]

[0037] in , is a random number within that interval.

[0038] S424: Introduce reflective learning to update the individual's position. Compare and reflect on the red fox's position before and after the update, and update its position based on the optimal result. The formula for reflective learning is as follows:

[0039]

[0040] in, It was the red fox that followed. Fox moved to its updated location. This is the red fox's starting position. It is a new position that arises from reflection and learning; , representing the learning factor.

[0041] S425: Encirclement of Prey – Local Search Phase. After spotting suspicious prey, the red fox hunts by observing its surroundings, camouflaging and concealing itself, and waiting for an opportune moment. The mathematical model for this is as follows:

[0042]

[0043] in, This is a random number within the specified interval, balancing the approximation and camouflage phases. The red fox's activity radius when surrounding its prey. The expression is as follows:

[0044]

[0045] in, , is defined as the approximation parameter; It is between and The random number between these values ​​represents the red fox's observation angle; The initial random number represents the influence of weather factors during a red fox's hunt. Once the time is right, the red fox begins to surround and approach its prey, and its position is updated as follows:

[0046]

[0047] in, It is between and Random numbers between; This indicates the new location for the red fox.

[0048] S426: Employ an elite-reverse learning strategy. Let the position of the elite individual in the current group be denoted as .

[0049]

[0050] Then the reverse solution Defined as:

[0051]

[0052] in, , yes Boundary values.

[0053] S427: Evade hunters and develop the population. The bottom 5% of individuals with the worst fitness are hunted and eliminated by hunters, and the population with the best fitness is selected. Red Fox Couple Establish a habitat model. The formula for calculating the habitat center is:

[0054]

[0055] in, , This represents the best-fitting red fox individual. The activity diameter of a fox pack in its habitat is defined as:

[0056]

[0057] S428: Determine if the termination condition is met. If the termination condition is met, obtain the optimal individual position, i.e., the optimal weights and threshold of the BP neural network, and then construct a new BP neural network.

[0058] S429: If the termination conditions are not met, the worst individual is eliminated, and new individuals are bred based on the best individuals, and then the process is changed to S422.

[0059] Preferably, in S424 of the improved Red Fox optimization algorithm, the reflective learning process... Apply constraints, specifically as follows:

[0060]

[0061] in: Random numbers on the screen For the current number of times, This represents the maximum number of iterations. Individuals exhibit strong reflective behavior in the early stages, enhancing their hunting abilities and achieving self-improvement; however, as individuals get closer to their prey, reflective behavior gradually decreases, and the learning factor... It also gradually became smaller.

[0062] Preferably, in S426, the elite reverse learning strategy selects the top 20% of individuals in terms of fitness to form an elite fox group. And the elite fox population was obtained using the inverse solution. In mixed fox packs Then select the top 50% of red foxes in terms of fitness as the next generation.

[0063] Preferably, in S427 of the improved red fox optimization algorithm, during the process of avoiding hunters and developing the population, parameters are set when developing the population. The details are as follows:

[0064]

[0065] when hour, When a pair of red foxes reproduce, they produce offspring, which can be denoted as:

[0066] .

[0067] Compared with the prior art, the beneficial effects of the present invention are:

[0068] 1. This invention establishes a demand prediction model for high-value medical consumables and spare parts based on a backpropagation (BP) neural network to predict the consumption of these parts. First, historical data on the demand for high-value medical consumables and spare parts is used as input to the prediction model. Then, an improved Red Fox optimization algorithm is used to optimize the parameters of the BP neural network, determining the optimal parameters to complete the model construction. By adding new sample data to the high-value medical consumables and spare parts demand prediction model for training, the accuracy of demand prediction for high-value medical consumables and spare parts is improved.

[0069] 2. This invention improves the performance of the Red Fox Optimization Algorithm by modifying it. First, reflective learning is introduced into the Red Fox Optimization Algorithm to compare the positions of the red foxes before and after updates, thereby enhancing the hunting ability of individuals and preventing the algorithm from getting trapped in local optima. Second, an elite back-learning strategy is adopted to select high-quality individuals as the next generation of the population, increasing population diversity, preventing premature convergence, and improving algorithm efficiency. Compared with other traditional optimization algorithms, the improved algorithm shows improvements in convergence speed and accuracy. Furthermore, by utilizing the Red Fox Optimization Algorithm for parameter optimization, it reduces the possibility of unsatisfactory prediction results caused by subjective parameter settings, providing more accurate prediction services for the demand forecasting of high-value medical consumables and spare parts. Attached Figure Description

[0070] Figure 1 This is a flowchart of the high-value medical consumables spare parts demand forecasting method in Embodiment 1 of the present invention;

[0071] Figure 2 This is a flowchart of the improved Red Fox optimization algorithm in Embodiment 1 of the present invention; Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Please see Figure 1-2 The present invention provides the following technical solutions:

[0074] A method for predicting the demand for high-value medical consumables spare parts based on an improved Red Fox optimization algorithm and a BP neural network, the method comprising the following steps:

[0075] S1: Obtain high-value medical consumable consumption data over a period of time, analyze the factors affecting spare parts, and normalize the data to map the original data to the [0,1] interval.

[0076] S2: Constructing a demand prediction model for high-value medical consumables spare parts based on BP neural network: Determine the network topology based on the number of input and output parameters, determine the number of network training iterations, training objectives, learning rate, and select the BP neural network transfer function and training method.

[0077] In this embodiment, constructing a demand prediction model for high-value medical consumables spare parts based on a BP neural network includes the following steps:

[0078] S21: Determine the number of network layers. Theoretically, it has been proven that a single-hidden-layer BP neural network can achieve arbitrary nonlinear mappings by appropriately increasing the number of neurons. In most cases, a single hidden layer is sufficient. This invention uses a single hidden layer.

[0079] S22: Analyze and determine the number of factors affecting the demand for high-value medical consumables spare parts, determine the number of neurons in the input layer and output layer of the BP neural network, and combine with S21 to determine the topology of the BP neural network.

[0080] The number of neurons in the hidden layer is determined empirically.

[0081]

[0082] in, This represents the number of neurons in the hidden layer. and These represent the number of neurons in the input layer and the output layer, respectively. yes The constant.

[0083] S23: Selection of transfer function and training method; hidden layer adopts... The output layer uses a linear function, and the commonly used gradient descent method is chosen for training.

[0084] S24: Determine the number of network training iterations, training objectives, and learning rate.

[0085] S25: Obtain training sample data and train the network.

[0086] S3: Initialize the weights and threshold of the BP neural network, setting them empirically as follows: .

[0087] S4: Obtain the optimal weights and thresholds of the BP neural network based on the improved Red Fox optimization algorithm.

[0088] In this embodiment, obtaining the optimal weights and thresholds of the BP neural network includes the following steps:

[0089] S41: Initialize the red fox population, and use the error between the spare parts prediction result and the actual value as the fitness function of the red fox algorithm.

[0090] S42: Select the position of the red fox individual with the best fitness value, simulate the red fox's life habits such as foraging, hunting, and escaping hunters according to the improved red fox optimization algorithm, and iteratively update the position of the red fox individual through its five stages of finding prey, reflecting and learning, attacking prey, elite reverse strategy, and escaping hunters and developing the population. Select the best individual after iteration until the maximum number of iterations is reached or the fitness requirement is met.

[0091] The improved Red Fox optimization algorithm includes the following steps:

[0092] S421: Algorithm Assumptions: The individual positions of the red fox population are initialized as follows: ,in The number of individuals in the population. , , , Let be the dimension of the individual's position; Indicates the first During the nth iteration The location of the red fox is determined. A fitness function is pre-defined, assuming the red foxes move within a specific spatial range. The red fox with the best fitness is recorded as the optimal individual, referred to as the "optimal individual". fox.

[0093] S422: Calculate the fitness value and select the red fox individual with the lowest fitness. The mean square error (MSE) between the predicted and actual values ​​in the BP neural network training set is chosen as the fitness function for the red fox individual.

[0094]

[0095] in This represents the actual demand for high-value medical consumables. To predict the demand for high-value medical consumables, the smaller the fitness value of a red fox individual, the higher the prediction accuracy of the algorithm.

[0096] S423: Hunting Prey – Global Search Phase. In this phase, individual red foxes venture out to hunt prey. The fox pack communicates with each other, sharing the easiest locations to catch prey, causing the pack to move towards the optimal individual's location. If the fitness value after moving is better than the original location, the location is updated; otherwise, it returns to the original location. In one iteration, the individual and the optimal individual... The distance can be expressed as:

[0097]

[0098] The formula for updating individual location is:

[0099]

[0100] in , is a random number within that interval.

[0101] S424: Introduce reflective learning to update the individual's position. Compare and reflect on the red fox's position before and after the update, and update its position based on the optimal result. The formula for reflective learning is as follows:

[0102]

[0103] in, It was the red fox that followed. Fox moved to its updated location. This is the red fox's starting position. It is a new position that arises from reflection and learning; , representing the learning factor.

[0104] Reflecting on learning Apply constraints, specifically as follows:

[0105]

[0106] in: Random numbers on the screen For the current number of times, Maximum number of iterations

[0107] By adopting the above technical solution, in order to prevent excessive randomness in the individual reflective learning process, the learning factors are... Imposing constraints initially leads to strong reflective behavior in individuals, enhancing their hunting abilities and achieving self-improvement. However, as individuals get closer to their prey, this reflective behavior gradually decreases, and the learning factor diminishes. It also gradually became smaller.

[0108] S425: Encirclement of Prey – Local Search Phase. After spotting suspicious prey, the red fox hunts by observing its surroundings, camouflaging and concealing itself, and waiting for an opportune moment. The mathematical model for this is as follows:

[0109]

[0110] in, This is a random number within the specified interval, balancing the approximation and camouflage phases. The red fox's activity radius when surrounding its prey. The expression is as follows:

[0111]

[0112] in, , is defined as the approximation parameter; It is between and The random number between these values ​​represents the red fox's observation angle; The initial random number represents the influence of weather factors during a red fox's hunt. Once the time is right, the red fox begins to surround and approach its prey, and its position is updated as follows:

[0113]

[0114] in, It is between and Random numbers between; This indicates the new location for the red fox.

[0115] S426: Employ an elite-reverse learning strategy. Let the position of the elite individual in the current group be denoted as .

[0116]

[0117] Then the reverse solution Defined as:

[0118]

[0119] in, , yes Boundary values.

[0120] In the elite reverse learning strategy, a certain proportion of the best individuals in the fox group is set at 20%, and a mixed fox group is selected. The first 50% of the red foxes will be used as the next generation.

[0121] By adopting the above technical solution, it is easier to select high-quality individuals as the next generation of the population, so as to effectively increase the diversity of the population, and at the same time, to a certain extent, prevent the "premature convergence" of the algorithm and improve the efficiency of the algorithm.

[0122] S427: Evade hunters and develop the population. The bottom 5% of individuals with the worst fitness are hunted and eliminated by hunters, and the population with the best fitness is selected. Red Fox Couple Establish a habitat model. The formula for calculating the habitat center is:

[0123]

[0124] in, , This represents the best-fitting red fox individual. The activity diameter of a fox pack in its habitat is defined as:

[0125]

[0126] While evading hunters and developing their population, parameters are set during population development. The details are as follows:

[0127]

[0128] when hour, When a pair of red foxes reproduce, they produce offspring, which can be denoted as:

[0129]

[0130] By adopting the above technical solution, parameters are set. This helps distinguish between the migration or breeding stages of fox packs.

[0131] S428: Determine if the termination condition is met. If the termination condition is met, obtain the optimal individual position, i.e., the optimal weights and threshold of the BP neural network, and then construct a new BP neural network.

[0132] S429: If the termination conditions are not met, the worst individual is eliminated, and new individuals are bred based on the best individuals, and then the process is changed to S422.

[0133] By adopting the above technical solution, firstly, reflective learning is introduced into the red fox optimization algorithm to compare the positions of red foxes before and after the update, so as to improve the hunting ability of individuals and avoid the algorithm from getting stuck in local optima; secondly, an elite reverse learning strategy is adopted to select high-quality individuals as the next generation of the population, so as to increase the diversity of the population, prevent the algorithm from converging too early, and improve the efficiency of the algorithm.

[0134] S43: If the termination condition is not met, eliminate the worst individual in the red fox population and use the best individual to reproduce offspring. Repeat step S42 until the iteration ends, and obtain the best individual position as the optimal weight and threshold of the BP neural network.

[0135] By adopting the above technical solution and using the improved Red Fox optimization algorithm to optimize the parameters of the BP neural network, the blindness of manually selecting parameters is avoided, thereby effectively predicting the demand for high-value medical consumables spare parts.

[0136] S5: Construct a new demand prediction model for high-value medical consumables spare parts based on the optimal weights and thresholds.

[0137] S6: Use the historical data of high-value medical consumables and spare parts consumption within a certain time period as the training set to train and test the high-value medical consumables and spare parts demand prediction model. After inverse normalization of the prediction results, output the prediction results.

[0138] This invention establishes a demand prediction model for high-value medical consumables and spare parts based on a backpropagation (BP) neural network to predict the consumption of these parts. First, historical consumption data is used as input to the prediction model. Then, an improved Red Fox optimization algorithm is combined to optimize the BP neural network parameters, determining the optimal parameters and thus completing the model construction. New sample data is added to the high-value medical consumables and spare parts demand prediction model for training, thereby improving the accuracy of demand prediction.

[0139] This invention improves the performance of the Red Fox Optimization Algorithm by modifying it. First, reflective learning is introduced into the Red Fox Optimization Algorithm to compare the positions of red foxes before and after updates, enhancing the hunting ability of individuals and preventing the algorithm from getting trapped in local optima. Second, an elite back-learning strategy is employed to select high-quality individuals for the next generation of the population, increasing population diversity, preventing premature convergence, and improving algorithm efficiency. Compared to other traditional optimization algorithms, the improved algorithm shows improvements in convergence speed and accuracy. Furthermore, by utilizing the Red Fox Optimization Algorithm for parameter optimization, it reduces the likelihood of unsatisfactory prediction results caused by subjective parameter settings, providing more accurate prediction services for the demand forecasting of high-value medical consumables and spare parts.

[0140] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process or method. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process or method that includes said element.

[0141] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting the demand for high-value medical consumables spare parts, characterized in that, The method includes the following steps: S1: Obtain high-value medical consumables consumption data over a period of time, analyze the factors affecting spare parts, and normalize the data to map the original data to the [0,1] interval; S2: Constructing a demand prediction model for high-value medical consumables spare parts based on BP neural network: Determine the network topology based on the number of input and output parameters, determine the number of network training iterations, training objectives, learning rate, and select the BP neural network transfer function and training method; S3: Initialize the weights and threshold of the BP neural network, setting them empirically as follows: ; S4: Obtain the optimal weights and thresholds of the BP neural network based on the improved Red Fox Optimization Algorithm; S41: Initialize the red fox population and use the error between the spare parts prediction result and the actual value as the fitness function of the red fox algorithm; S42: Select the position of the red fox individual with the best fitness value, simulate the red fox's life habits of foraging, hunting, and escaping hunters according to the improved red fox optimization algorithm, and iteratively update the position of the red fox individual through its five stages of finding prey, reflecting and learning, surrounding prey, elite reverse strategy, escaping hunters and developing the population. Select the best individual after iteration until the maximum number of iterations is reached or the fitness requirement is met. Reflective learning is introduced to update individual positions. The positions of the red fox before and after the update are compared and the best position is selected for updating. To prevent excessive randomness in the reflective learning process, constraints are imposed on the learning factors. Individuals have strong reflective behavior in the early stage, but the reflective behavior gradually decreases as the individual gets closer to the prey. An elite reverse learning strategy is adopted to select high-quality individuals as the next generation of the population. S43: If the termination condition is not met, eliminate the worst individual in the red fox population and use the best individual to reproduce offspring. Repeat step S42 until the iteration ends, and obtain the best individual position as the optimal weight and threshold of the BP neural network prediction model. S5: Construct a new demand forecasting model for high-value medical consumables spare parts based on optimal weights and thresholds; S6: Use the historical data of high-value medical consumables and spare parts consumption within a certain time period as the training set to train and test the high-value medical consumables and spare parts demand prediction model. After inverse normalization of the prediction results, output the prediction results.

2. The method for predicting the demand for high-value medical consumables spare parts according to claim 1, characterized in that: Step S2, which involves constructing a demand prediction model for high-value medical consumables and spare parts based on a BP neural network, specifically includes the following steps: S21: Determine the number of network layers, using a single hidden layer; S22: Analyze and determine the number of factors affecting the demand for high-value medical consumables spare parts, determine the number of neurons in the input layer and output layer of the BP neural network, and combine with S21 to determine the topology of the BP neural network; The number of neurons in the hidden layer is determined empirically. in, This represents the number of neurons in the hidden layer. and These represent the number of neurons in the input layer and the output layer, respectively. yes The constant; S23: Selection of transfer function and training method; hidden layer adopts... The function used in the output layer is a linear function, and the training method chosen is gradient descent. S24: Determine the number of network training iterations, training objectives, and learning rate; S25: Obtain training sample data and train the network.

3. The method for predicting the demand for high-value medical consumables spare parts according to claim 2, characterized in that: The improved Red Fox optimization algorithm in step S42 specifically includes the following steps: S421: Algorithm Assumptions: The individual positions of the red fox population are initialized as follows: ,in The number of individuals in the population. , , , Let be the dimension of the individual's location; pre-define the fitness function, assuming the red fox's activity is within a specific spatial range, and denote the red fox with the best fitness as the optimal individual, referred to as . fox; S422: Calculate the fitness value and select the red fox individual with the lowest fitness; select the mean square error (MSE) between the predicted and actual values ​​in the BP neural network training set as the fitness function for the red fox individual: in This represents the actual demand for high-value medical consumables. For predicting the demand for high-value medical consumables, the smaller the fitness value of a red fox individual, the higher the prediction accuracy of the algorithm. S423: Hunting Prey – Global Search Phase; In this phase, individual red foxes venture out to hunt prey, and the fox pack communicates with each other, sharing the locations where prey is most easily caught. The pack moves towards the location of the optimal individual. If the fitness value after the move is better than the original location, the location is updated; otherwise, it returns to the original location. In one iteration, the individual and the optimal individual... The distance is expressed as: The formula for updating individual location is: in , is a random number within that interval; Indicates the first During the nth iteration The location of only the red fox; S424: Introduce reflective learning to update individual positions; compare and contrast the red fox's positions before and after the update, and select the optimal position to update. The formula for reflective learning is as follows: in, It was the red fox that followed. Fox moved to its updated location. It is a new position that arises from reflection and learning; , representing the learning factor; S425: Encirclement of Prey – Local Search Phase; After discovering suspicious prey, the red fox hunts by observing its surroundings, camouflaging and hiding, and waiting for an opportunity to strike. The mathematical model is as follows: in, The random number within this interval is used to balance the approximation and camouflage phases; the red fox's activity radius when surrounding its prey. The expression is as follows: in, , is defined as the approximation parameter; It is between and The random number between these values ​​represents the red fox's observation angle; The initial random number represents the influence of weather factors during a red fox's hunt; once the time is right, the red fox begins to surround and approach its prey, and its position is updated as follows: in, It is between and Random numbers between; This indicates the new location for the red fox; S426: Adopt an elite-reverse learning strategy; let the position of the elite individual in the current group be... Then the reverse solution Defined as: in, , yes Boundary values; S427: Evade hunters and develop the population; the bottom 5% of individuals with the worst fitness are hunted and eliminated by hunters, and the population with the best fitness is selected. Red Fox Couple Establish a habitat model; the formula for calculating the habitat center is: in, , This represents the best-fitting red fox individual; the habitat fox pack activity diameter is defined as: S428: Determine if the termination condition is met; if the termination condition is met, obtain the optimal individual position, i.e., the optimal weights and thresholds of the BP neural network, and then construct a new BP neural network. S429: If the termination conditions are not met, the worst individual is eliminated, and new individuals are bred based on the best individuals, and then the process is changed to S422.

4. The method for predicting the demand for high-value medical consumables spare parts according to claim 3, characterized in that: To prevent excessive randomness during reflective learning, the following constraints are imposed on the learning factor in S424: in: Random numbers on the screen For the current number of times, This represents the maximum number of iterations; individuals exhibit strong reflective behavior in the early stages, enhancing their hunting abilities and achieving self-improvement; however, as individuals get closer to their prey, reflective behavior gradually decreases, and the learning factor... It also gradually became smaller.

5. The method for predicting the demand for high-value medical consumables spare parts according to claim 3, characterized in that: In S426, the elite reverse learning strategy selects the top 20% of individuals by fitness to form an elite fox group. And the elite fox population was obtained using the inverse solution. In mixed fox populations Then select the top 50% of red foxes in terms of fitness as the next generation.

6. The method for predicting the demand for high-value medical consumables spare parts according to claim 3, characterized in that: In S427, during the process of evading hunters and developing the population, parameters are set when developing the population. The details are as follows: when hour, When the red fox couple reproduces, new individuals are produced, and these new individuals are denoted as: Setting parameters This helps distinguish between the migration or breeding stages of fox populations.

Images