LKAN neural network optimization method based on mouse swarm algorithm, equipment support system evaluation method and electronic equipment

Abstract

The embodiment of the invention relates to the technical field of data processing based on a specific model, in particular to an LKAN neural network optimization method based on a mouse swarm algorithm, an equipment support system evaluation method and electronic equipment, and the method comprises the steps: determining an individual candidate solution and a current optimal solution, and updating the individual candidate solution through the current optimal solution; performing iterative operation, evaluating the current optimal solution through an objective function, generating an improved mouse swarm algorithm, and returning a global optimal solution; and constructing a Legendre polynomial neural network, and setting an input layer of the Legendre polynomial neural network. According to the equipment support system efficiency evaluation method provided by the embodiment of the invention, the efficiency evaluation of the equipment support system is realized based on the improved LKAN neural network, and the actual effect of the current equipment support system in completing the equipment support task under the existing equipment support capability can be evaluated; the method is comprehensive embodiment of the effective degree of the equipment support activity to complete the specified task.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of data processing based on a specific model, and in particular, to an optimization method for an LKAN neural network based on a mouse swarm algorithm, an evaluation method for an equipment support system, and an electronic device. Background Art

[0002] The equipment support system is an important logistics support for the troops, which can provide sufficient combat resources for the troops to ensure the smooth progress of combat. Conducting an effectiveness evaluation of the equipment support system can effectively discover the problems existing in the system and timely formulate improvement plans to improve the effectiveness of the equipment support system, which is one of the key technologies in the field of equipment support. Therefore, how to effectively and accurately evaluate the effectiveness of the equipment support system has become a research hotspot. At present, researchers at home and abroad mainly use the analytic hierarchy process, ADC model, fuzzy comprehensive evaluation method, and neural network model for effectiveness evaluation. The analytic hierarchy process is based on expert experience and realizes the quantitative analysis of decision-making results through a judgment matrix. However, this method follows principles such as top-down and layer-by-layer transmission, which is not conducive to reflecting the feedback effect of the index layer on the target layer and the mutual influence between various indexes at different levels, and has strong subjective factors. The ADC model represents the system effectiveness using the functions of the system availability vector A, credibility matrix D, and inherent ability matrix C. However, when the evaluation object includes multiple sub-objects and each sub-object includes multiple initial states, the complexity of determining the initial state and the complexity of calculating the state transition probability of this model will increase exponentially. The fuzzy comprehensive evaluation method is based on fuzzy mathematics and improves the accuracy and credibility of the evaluation results. However, when the number of system effectiveness evaluation indexes increases, the relative membership coefficient will be relatively small, which may lead to the inability of the weight vector to match the fuzzy matrix and cause the evaluation to fail. Using a neural network for effectiveness evaluation can effectively solve the problem of subjective factor influence in the evaluation process and can generate an accurate and objective intelligent evaluation model. Based on the above analysis, it can be seen that the effectiveness evaluation methods of the equipment support system in related technologies have problems such as relying on expert experience and strong subjectivity, resulting in a certain degree of subjectivity in the evaluation results.

[0003] In view of the above problems, the related technologies have not yet proposed effective technical solutions, which can no longer meet people's requirements and urgently need to be improved. Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to provide an optimization method for an LKAN neural network based on a mouse swarm algorithm, an evaluation method for an equipment support system, and an electronic device, so as to solve the problems of relying on expert experience and strong subjectivity in related technologies.

[0005] According to one aspect of the embodiments of the present application, an optimization method for a LKAN neural network based on a mouse swarm algorithm is provided, including: determining an individual candidate solution and a current optimal solution, and updating the individual candidate solution through the current optimal solution; performing iterative operations, evaluating the current optimal solution through an objective function, adjusting the individual candidate solution and the current optimal solution according to the relationship between the individual candidate solution and the search space, generating an improved mouse swarm algorithm and returning the global optimal solution; constructing a Legendre polynomial neural network, and setting the input layer of the Legendre polynomial neural network according to the global optimal solution.

[0006] According to at least one implementation manner of the embodiments of the present application, the determining an individual candidate solution and a current optimal solution, and updating the individual candidate solution through the current optimal solution further includes: initializing the mouse swarm algorithm, setting the current individual position, the current best position, the maximum number of iterations, the exploration parameter, and the exploitation parameter in the position update formula; balancing the exploration and exploitation mechanisms of the mouse swarm algorithm through the exploration parameter and the exploitation parameter; subtracting the current individual position from the current best position, taking the absolute value of the subtraction result, and obtaining the next individual position.

[0007] According to at least one implementation manner of the embodiments of the present application, the performing iterative operations, evaluating the current optimal solution through an objective function, adjusting the individual candidate solution and the current optimal solution according to the relationship between the individual candidate solution and the search space, generating an improved mouse swarm algorithm and returning the global optimal solution further includes: initializing the parameters of the position update formula in the mouse swarm algorithm, where the parameters include the population size, the maximum number of iterations, and the current global optimal solution; detecting the current individual position in the position update formula of the mouse swarm algorithm, and if the current individual position exceeds the search space of the mouse swarm algorithm, reassigning the current individual position back to the previous position.

[0008] According to at least one implementation manner of the embodiments of the present application, in the process of reassigning the current individual position back to the previous position, evaluating the individual candidate solution through the objective function includes: initializing the candidate solution set corresponding to the individual candidate solution and the objective function, calculating the fitness value of the individual candidate solution through the objective function, and updating the position of the individual candidate solution; dynamically adjusting the parameters in the position update formula according to the value of the objective function to keep the individual candidate solution within a preset search space; repeating the above steps until the number of iterations reaches the preset maximum number of iterations or the objective function meets the convergence condition.

[0009] According to at least one implementation manner of the embodiments of the present application, when constructing the Legendre polynomial neural network, the input layer of the Legendre polynomial neural network is set according to the global optimal solution, specifically: the Legendre polynomial neural network includes an input layer, a hidden layer, and an output layer, the input layer is used to receive the global optimal solution, the hidden layer uses the Legendre polynomial as a basis function, and the output layer is used to output the optimization result.

[0010] According to another aspect of the embodiments of the present application, there is provided an equipment support system evaluation method, including the LKAN neural network optimization method based on the mouse swarm algorithm, and further including: acquiring data samples, preprocessing the data samples to generate preprocessed data; setting the Legendre polynomial neural network, and corresponding the number of input layers of the Legendre polynomial neural network to the number of equipment support system effectiveness evaluation indicators; training the Legendre polynomial neural network by using the improved mouse swarm algorithm, and evaluating the effectiveness of the equipment support system by using the trained Legendre neural network to obtain the equipment support system effectiveness evaluation value.

[0011] According to at least one implementation manner of the embodiments of the present application, when acquiring data samples and preprocessing the data samples to generate preprocessed data, specifically: performing normalization processing on the data samples to generate normalized data, and the normalized data is used to eliminate the influence of the index unit and its numerical order of magnitude on the data samples.

[0012] According to at least one implementation manner of the embodiments of the present application, further setting the Legendre polynomial neural network includes: using the Legendre polynomial as a basis function in the output layer of the Legendre polynomial neural network; training the Legendre polynomial neural network based on the improved mouse swarm algorithm, initializing the individual positions of the position update formula in the mouse swarm algorithm, so that the individual positions correspond to the combination of the parameters of the Legendre polynomial neural network; evaluating the individual positions of the position update formula according to the objective function, and updating the corresponding individual positions according to the evaluation results; repeating the above steps until the number of iterations reaches the maximum number of iterations.

[0013] According to still another aspect of the embodiments of the present application, there is provided an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the method.

[0014] According to yet another aspect of the embodiments of the present application, there is provided a computer program product, including a computer program, and when the computer program is executed by a processor, the method is implemented.

[0015] The beneficial technical effects of the embodiments of the present application are as follows:

[0016] By introducing a new parameter adjustment mechanism, the embodiments of the present application improve the dynamic balance in the search process of the mouse swarm algorithm, and enhance the search efficiency and global optimization ability of the algorithm. The improved mouse swarm algorithm is applied to the parameter adjustment of the neural network. Instead of using the standard basis functions and spline functions of the related technology, the Legendre polynomials are used. The Legendre polynomials are a series of orthogonal polynomials, which play a crucial role in the approximation theory and numerical analysis. The embodiments of the present application use the improved mouse swarm algorithm to dynamically adjust the network parameters of the Legendre neural network (LKAN), optimize the neural network, and use the improved mouse swarm algorithm to adjust the parameters of the LKAN neural network. The position of the mouse corresponds to a combination of the parameters of the LKAN neural network.

[0017] The equipment support system effectiveness evaluation method provided by the embodiments of the present application is based on the improved LKAN neural network, realizes the evaluation of the equipment support system effectiveness, and can evaluate the actual effect exerted when the current equipment support system completes the equipment support task under the existing equipment support capabilities. It is a comprehensive reflection of the effective degree of the equipment support activities to complete the specified tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific implementation manners of the embodiments of the present application or the technical solutions in the related art, the following will briefly introduce the drawings required for use in the description of the specific implementation manners or the related technology. Obviously, the drawings in the following description are only some implementation manners of the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a flowchart of the LKAN neural network optimization method based on the mouse swarm algorithm.

[0020] Figure 1A It is a system architecture diagram of the LKAN neural network optimization system based on the mouse swarm algorithm.

[0021] Figure 2 It is a flowchart of the optimization technical solution provided from step S11 to step S13.

[0022] Figure 3 It is a flowchart of the optimization technical solution provided from step S21 to step S22.

[0023] Figure 4 It is a flowchart of the optimization technical solution provided from step S221 to step S223.

[0024] Figure 5It is the flowchart of the evaluation method for the equipment support system.

[0025] Figure 5A It is the architecture diagram of the equipment support system evaluation system.

[0026] Figure 6 It is the flowchart of the method for optimizing the LKAN neural network using the improved mouse swarm algorithm.

[0027] Figure 7 It is the flowchart of the method for evaluating the effectiveness of the equipment support system based on the optimization of the LKAN neural network.

[0028] Figure 8 It is the structural schematic diagram of the electronic device. Detailed implementation manners

[0029] In order to enable those skilled in the art to better understand the solution of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the embodiments of the present application.

[0030] It should be noted that, without conflict, the embodiments in the embodiments of the present application and the features in the embodiments can be combined with each other. The embodiments of the present application will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0031] The equipment system effectiveness evaluation method in the related technology is highly subjective, relying on expert experience, and it is difficult to accurately evaluate the effectiveness of the equipment support system. The Legendre polynomial Kolmogorov - Arnold neural network (KAN) optimized by the improved mouse swarm algorithm (MRSO) provided in the embodiments of the present application can handle the difficult - to - analyze regularities within the system, has strong generalization ability and fast learning convergence speed, and can well avoid the interference of human factors, and can evaluate the effectiveness of the equipment support system more objectively and accurately.

[0032] As Figure 1 shown, the embodiments of the present application provide an LKAN neural network optimization method based on the mouse swarm algorithm, which is applied to the LKAN neural network optimization system based on the mouse swarm algorithm. The LKAN neural network optimization method here can be applied to the evaluation of the effectiveness of the equipment support system. The method steps include:

[0033] Step S1, determine the individual candidate solution and the current optimal solution, and update the individual candidate solution through the current optimal solution.

[0034] Step S2: Perform iterative operations, evaluate the current optimal solution through the objective function, adjust the individual candidate solution and the current optimal solution according to the relationship between the individual candidate solution and the search space, generate an improved rat swarm algorithm, and return the global optimal solution.

[0035] Step S3: Construct a Legendre polynomial neural network, and set the input layer of the Legendre polynomial neural network according to the global optimal solution.

[0036] For the method steps disclosed in the above embodiments, for the purpose of simple description, the method steps are expressed as a series of action combinations. However, those skilled in the art should know that the embodiments of the present invention are not limited by the described action sequence, because according to the embodiments of the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0037] Principle of rat swarm algorithm:

[0038] (1) Chasing prey: The chasing behavior of rats is a typical social activity. The most effective search agent is determined to be the rat that knows the location of the prey. The others adjust their positions according to the position of this best rat as follows:

[0039]

[0040] Where, represents the position of the i-th rat (solution), and (t) represents the current iteration number. Represents the position of the best candidate solution found so far. The calculation of (A) is as follows:

[0041]

[0042] R and C are random values, where the range of R is between [1, 5], and the range of C is between [0, 2]. These values are used as parameters for the exploration and exploitation mechanisms in the algorithm: R = rand(1, 5), C = rand(0, 2).

[0043] (2) Fighting with the prey

[0044] The fighting behavior is mathematically expressed in the following way:

[0045]

[0046] The next position of the rat numbered i is denoted as The parameters A and C are crucial for balancing the exploration and exploitation mechanisms. A small value of A (e.g., 1) combined with a medium value of C emphasizes exploitation, while other values may shift the focus towards exploration.

[0047] (3) Modified Rat Swarm Optimization Algorithm (MRSO)

[0048] The Modified Rat Swarm Optimization Algorithm (MRSO) aims to improve the performance of RSO by balancing exploration and exploitation. This balance is achieved through the function in the "chasing prey" section, which is static and depends on a single parameter C. In MRSO, this function is reformulated as follows:

[0049]

[0050] F 3 = 2 × rand(0,1) - 1 × rand(0,1)

[0051] A Modified = F 1 × F 2 × F 3 , t = 0, 1, 2, …, Max iteration

[0052] The formula is updated as follows:

[0053]

[0054] After initializing the parameters A Modified , C, and R, the results are evaluated using the objective function, where the best solution is saved as Then, the positions of the rats are updated using the equation, and the parameters R, A, and A Modified are updated. If a rat's position exceeds the search space, it is reallocated to the previous center for adjustment. Then, the position of each rat is tested and evaluated using the objective function. If a better solution than is found, then is updated to this new best position. This process continues through the maximum number of iterations (Max iteration). Finally, the identified best position is used to select the position. These modifications improve the performance by obtaining the best fitness function.

[0055] 2. Kolmogorov - Arnold Network:

[0056] The LKAN architecture can improve the accuracy and efficiency of prediction. KAN follows the Kolmogorov - Arnold representation theorem, which is a fundamental result in approximation theory stating that any continuous multivariate function on a bounded domain can be represented as a superposition of univariate functions and linear operations. Mathematically, for a continuous function on the n - dimensional unit hypercube This theorem guarantees the existence of continuous univariate functions Φ j and φ ij such that

[0057]

[0058] where \(x=(x 1 ,x 2 ,\cdots,x n )\) is the input vector. According to this theorem, KAN is developed as a new type of neural network architecture, which approximates multivariate functions by combining univariate functions across layers. A general KAN consists of \(L\) layers. Given the input vector The output of KAN can be expressed as:

[0059]

[0060] where each \(\varPhi l represents a KAN layer with an \(n l -dimensional input and an \(n l+1 -dimensional output, and can be defined as a matrix of 1-dimensional functions:

[0061] \varPhi l =) l,ij ,l = 1,\cdots,L,i = 1,\cdots,n l ,j = 1,\cdots,n l+1 \cdots\cdots(3)

[0062] In the initial implementation of KAN, \(\varphi\) is defined as a weighted combination of the basis function \(b\) and the B-spline function,

[0063] \varphi(x)=\omega b b(x)+\omega s spline(x)\cdots\cdots(4)

[0064] where \(b(x)\) and \(spline(x)\) are defined as:

[0065]

[0066] where \(c i ,\omega b and \(\omega s are trainable parameters, denoted by \(\theta KAN =\{c i ,\omega b ,\omega s \}\). The B-spline i is characterized by the polynomial order \(k\) and the number of grid points \(g\).

[0067] In the embodiments of this application, KAN uses Legendre polynomials instead of using standard basis functions and spline functions. Legendre polynomials are a series of orthogonal polynomials that play a crucial role in approximation theory and numerical analysis. They can be defined on the interval \([-1,1]\) that satisfies the recurrence relation:

[0068]

[0069] Therefore, for LKAN, we have:

[0070] φ(x) = ∑ i c i T i (x)……(7)

[0071] where c i is a learnable parameter. KAN and LKAN have a similar structure in approximating complex functions, but they differ in the number of parameters and the dependence on grid points. In KAN, the total number of trainable parameters is affected by the input size, the number of hidden layers, the number of neurons in each layer, the grid size, and the polynomial order, and is represented by |θ| KAN ~O(n l H 2 (k + g)), where g is the grid size and k is the polynomial order. In contrast, LKAN eliminates the need for grid points, reducing the number of parameters to |θ| lKAN ~O(n l H 2 k), which simplifies the model and improves efficiency by only focusing on the polynomial order k.

[0072] In the present invention, LKAN is used to predict the effectiveness evaluation value of the equipment support system. The network takes relevant variables as inputs and predicts the effectiveness evaluation value of the equipment support system. To facilitate training the network, the total cost function CF is defined as the individual cost functions of the network, i.e.:

[0073] CF = MSE(y, y * )……(8)

[0074] where MSE represents the mean squared error and y * represents the true data. The LKAN architecture uses Legendre polynomials as basis functions, providing a more flexible and adaptive method to represent the solution space. This flexibility enables LKAN to capture complex patterns with higher accuracy and efficiency.

[0075] The technical solutions provided in steps S1 to S3 optimize the Legendre polynomial neural network based on the mouse swarm algorithm. Its main application scenario can be used for the evaluation of the effectiveness of the equipment support system, enabling the neural network to approximate complex functional relationships more efficiently, improving the accuracy and objectivity of the evaluation of the effectiveness of the equipment support system. By improving the mouse swarm algorithm to dynamically adjust the LKAN network parameters, the network can approximate complex functional relationships more efficiently, improve the prediction accuracy of the model, avoid the algorithm falling into local optima, enhance the global search ability of the algorithm, improve the optimization performance of the neural network, and find the global optimal solution within fewer iterations by reasonably setting parameters and optimization strategies, thereby improving the convergence speed. Using the Legendre polynomial as the basis function of the neural network can effectively handle complex non-linear relationships, enabling the LKAN network to perform well in dealing with complex problems. In summary, the technical solutions provided in steps S1 to S3 optimize the LKAN neural network parameters through the improved mouse swarm algorithm, not only improving the optimization accuracy and convergence speed of the network, but also enhancing the global search ability of the algorithm and the ability to adapt to complex problems, and having a good improvement effect on the evaluation results of the equipment insurance system effectiveness.

[0076] As Figure 1A shown, the embodiment of the present application provides an LKAN neural network optimization system based on the mouse swarm algorithm for implementing the LKAN neural network optimization method based on the mouse swarm algorithm, including:

[0077] An individual candidate solution update module, used to determine the individual candidate solution and the current optimal solution, and update the individual candidate solution through the current optimal solution;

[0078] A global optimal solution generation module, which performs iterative operations, evaluates the current optimal solution through the objective function, adjusts the individual candidate solution and the current optimal solution according to the relationship between the individual candidate solution and the search space, generates an improved mouse swarm algorithm and returns the global optimal solution;

[0079] A neural network input layer setting module, which constructs a Legendre polynomial neural network and sets the input layer of the Legendre polynomial neural network according to the global optimal solution.

[0080] The implementation manners of the systems described above are merely illustrative. For example, each functional module, unit or subsystem in the system may or may not be physically separated, or may or may not be physical units, that is, they may be located in the same place or distributed to multiple different systems and their subsystems or modules. Those skilled in the art can select some or all of the functional modules, units or subsystems according to actual needs to achieve the purpose of the embodiments of the present invention. For the above situations, those of ordinary skill in the art can understand and implement them without creative efforts.

[0081] AsFigure 2 As shown in the figure, in step S1, individual candidate solutions and the current optimal solution are determined, and the individual candidate solutions are updated through the current optimal solution, which further includes:

[0082] Step S11, initialize the mouse swarm algorithm, and set the current individual position, current best position, maximum number of iterations, exploration parameter, and exploitation parameter in the position update formula.

[0083] Step S12, balance the exploration and exploitation mechanisms of the mouse swarm algorithm through the exploration parameter and exploitation parameter.

[0084] Step S13, subtract the current individual position from the current best position, take the absolute value of the subtraction result, and obtain the next individual position.

[0085] In the optimization technical solution provided by steps S11 to S13, the LKAN network parameters are dynamically adjusted through the improved mouse swarm algorithm, enabling the network to approximate complex functional relationships more efficiently, improving the prediction accuracy of the model. By reasonably setting parameters and optimization strategies, the algorithm can find the global optimal solution within fewer iterations, improving the convergence speed. The Legendre polynomial, as the basis function, can effectively handle complex non-linear relationships, enabling the LKAN network to perform excellently in dealing with complex problems. By optimizing the algorithm, unnecessary calculations are reduced, improving the computational efficiency and reducing the computational cost. In summary, the optimization technical solution provided by steps S11 to S13 uses the improved mouse swarm algorithm to optimize the LKAN neural network parameters, not only improving the optimization accuracy and convergence speed of the network, but also enhancing the global search ability of the algorithm and the ability to adapt to complex problems.

[0086] As Figure 3 shown, in step S2, iterative operations are performed, the current optimal solution is evaluated through the objective function, and the individual candidate solutions and the current optimal solution are adjusted according to the relationship between the individual candidate solutions and the search space, generating an improved mouse swarm algorithm and returning the global optimal solution, which further includes:

[0087] Step S21, initialize the parameters of the position update formula in the mouse swarm algorithm, and the parameters include population size, maximum number of iterations, and the current global optimal solution.

[0088] Step S22, detect the current individual position in the position update formula of the mouse swarm algorithm. If the current individual position exceeds the search space of the mouse swarm algorithm, then reassign the current individual position back to the previous position. The purpose of step S22 is to retrieve whether the current individual position exceeds the search space, and the search space refers to the set of all possible solutions explored by the algorithm during the optimization process.

[0089] In the optimized technical solution provided from step S21 to step S22, the prediction accuracy of the mouse swarm algorithm model is improved, the algorithm is prevented from falling into local optimum, the global search ability of the algorithm is enhanced, and the optimization performance is improved. By reasonably setting parameters and optimization strategies, the algorithm can find the global optimum solution within fewer iteration times, improving the convergence speed. This not only improves the optimization accuracy and convergence speed of the network, but also enhances the global search ability of the algorithm and its ability to adapt to complex problems. It can be seen that as a basis function, the Legendre polynomial can effectively handle complex non-linear relationships, enabling the LKAN network to perform excellently when dealing with complex problems.

[0090] As Figure 4 shown, in step S22, during the process of reassigning the current individual position back to the previous position, the individual candidate solutions are evaluated through the objective function, including:

[0091] Step S221, initialize the candidate solution set corresponding to the individual candidate solutions and the objective function. Calculate the fitness value of the individual candidate solutions through the objective function, and update the positions of the individual candidate solutions.

[0092] Step S222, dynamically adjust the parameters in the position update formula according to the value of the objective function to keep the individual candidate solutions within the preset search space.

[0093] Step S223, repeat the above steps until the iteration times reach the preset maximum iteration times or the objective function meets the convergence condition.

[0094] In the optimized technical solution provided from step S221 to step S223, the candidate solution set refers to the set of all possible solutions, which are evaluated and updated during the optimization process. The initialization and update of the candidate solution set are the core parts of the algorithm, directly affecting the performance and optimization results of the algorithm. The objective function is a function used to evaluate the quality of each candidate solution. During the optimization process, the value of the objective function determines the fitness of the candidate solution, thus guiding the search direction of the algorithm. By evaluating the fitness value of the candidate solutions through the objective function, the algorithm can determine which solutions are closer to the global optimum solution. In the optimized technical solution provided from step S221 to step S223, by reasonably initializing the candidate solution set and the objective function, and dynamically adjusting the parameters in the position update formula, not only the optimization accuracy and convergence speed are improved, but also the global search ability of the algorithm and its ability to adapt to complex problems are enhanced, providing strong data processing ability and algorithm guarantee for the evaluation of the effectiveness of the equipment support system.

[0095] In step S3, construct a Legendre polynomial neural network, and set the input layer of the Legendre polynomial neural network according to the global optimum solution, specifically:

[0096] The Legendre polynomial neural network includes an input layer, a hidden layer, and an output layer. The input layer is used to receive the global optimal solution. The hidden layer uses the Legendre polynomial as the basis function. The output layer is used to output the optimization result.

[0097] In step S3, the single-layer structure of the LKAN network is used to reduce the computational complexity of the algorithm, which is applicable to various optimization problems. The parameters of the LKAN network are dynamically adjusted based on the improved rat swarm algorithm, so that the individual positions of the rats in the position update formula correspond to a combination of the parameters of the LKAN neural network. The prediction accuracy and efficiency of the model are improved by adjusting the parameters of the LKAN neural network through the optimization algorithm.

[0098] As Figure 5 shown in the equipment support system evaluation method, the equipment support system evaluation method is based on the LKAN neural network optimization method of the rat swarm algorithm, including:

[0099] Step T1, obtain data samples, preprocess the data samples to generate preprocessed data.

[0100] In step T1, the data samples refer to the data samples related to the equipment support system. The preprocessed data can be normalized data processing. Normalized data processing is part of the standardized data (Standardized Data). Other standardized data include Z-Score standardization (Standardization), Min-Max standardization, Log transformation, and Box-Cox transformation, etc. By preprocessing the data samples related to the equipment support system, the data is made more suitable for subsequent analysis, improving the performance and effect of the model. The data is cleaned, extracted, selected, and transformed in features, and the data dimension can be reduced by methods such as principal component analysis.

[0101] Step T2, set the Legendre polynomial neural network, and correspond the number of input layers of the Legendre polynomial neural network to the number of equipment support system effectiveness evaluation indicators.

[0102] Step T3, use the improved rat swarm algorithm to train the Legendre polynomial neural network, and use the trained Legendre neural network to evaluate the effectiveness of the equipment support system to obtain the equipment support system effectiveness evaluation value.

[0103] The technical solutions provided by steps T1 to T3 achieve an efficient evaluation of the effectiveness of the equipment support system by preprocessing the relevant data of the equipment support system, constructing and training a neural network model. By normalizing, cleaning, extracting, selecting, and reducing the dimensions of the data (such as principal component analysis), the data is made more suitable for subsequent analysis, improving the data quality. The number of input layers of the Legendre polynomial neural network is corresponding to the number of effectiveness evaluation indicators of the equipment support system, ensuring that the model can accurately process the data related to the equipment support system and improving the adaptability of the model. The improved rat swarm algorithm is used to train the neural network, improving the efficiency and convergence speed of the training process. Finally, the effectiveness of the equipment support system is evaluated by the trained Legendre polynomial neural network, and an accurate effectiveness evaluation value can be obtained, providing a reliable basis for the optimization and improvement of the equipment support system.

[0104] The technical solutions provided by steps T1 to T3 can be combined with the technical solutions provided by steps S1 to S3, or can be combined with the optimized technical solutions of any one or more steps in the technical solutions of steps S1 to S3 to form various technical solutions. Steps S1 to S3 optimize the LKAN neural network based on the improved rat swarm algorithm and apply the optimization results to steps T1 to T3 to achieve the evaluation of the effectiveness of the equipment support system.

[0105] Such as Figure 5A As shown, the embodiment of the present application provides an equipment support system evaluation system for implementing the equipment support system evaluation method, including:

[0106] A preprocessing data generation module for obtaining data samples, preprocessing the data samples, and generating preprocessed data.

[0107] A neural network setting module for setting the Legendre polynomial neural network and corresponding the number of input layers of the Legendre polynomial neural network to the number of effectiveness evaluation indicators of the equipment support system.

[0108] An equipment support system effectiveness evaluation module for training the Legendre polynomial neural network using the improved rat swarm algorithm and evaluating the effectiveness of the equipment support system using the trained Legendre neural network to obtain the equipment support system effectiveness evaluation value.

[0109] The implementation manners of the system described above are merely illustrative. For example, each functional module, unit, or subsystem in the system may or may not be physically separated, or may or may not be physical units. That is, they may be located in the same place or distributed to multiple different systems and their subsystems or modules. Those skilled in the art can select some or all of the functional modules, units, or subsystems according to actual needs to achieve the purpose of the embodiments of the present invention. For the above situations, those of ordinary skill in the art can understand and implement them without creative efforts.

[0110] Preferably, in step T1, data samples are obtained, and the data samples are preprocessed to generate preprocessed data. Specifically, the data samples are normalized to generate normalized data, and the normalized data is used to eliminate the influence of the index unit and its numerical order of magnitude on the data samples.

[0111] Preferably, in step T2, the Legendre polynomial neural network is set up, which further includes:

[0112] Step T21, in the output layer of the Legendre polynomial neural network, the Legendre polynomial is used as the basis function.

[0113] Step T22, the Legendre polynomial neural network is trained based on the improved rat swarm algorithm. The individual positions in the position update formula of the rat swarm algorithm are initialized so that the individual positions correspond to the combinations of the parameters of the Legendre polynomial neural network.

[0114] Step T23, the individual positions in the position update formula are evaluated according to the objective function, and the corresponding individual positions are updated according to the evaluation results.

[0115] Step T24, the above steps are repeated until the number of iterations reaches the maximum number of iterations.

[0116] In the optimized technical solutions provided in steps T21 to T24, the setting and training process of the Legendre polynomial neural network are further refined. By using the Legendre polynomial as the basis function in the output layer and combining the improved rat swarm algorithm for parameter optimization, the efficient training and optimization of the equipment support system effectiveness evaluation model are realized. Finally, by optimizing the neural network structure and training algorithm, the performance and evaluation accuracy of the model are improved. The optimized technical solutions provided in steps T21 to T24 introduce the Legendre polynomial as the basis function in the output layer of the Legendre polynomial neural network. Utilizing its orthogonality and good approximation ability, the fitting ability and generalization performance of the model for complex data are enhanced. By initializing the positions of individuals in the rat swarm algorithm and corresponding them to the combinations of neural network parameters, a reasonable starting point is provided for the optimization process. The quality of individual positions is evaluated based on the objective function, and the individual positions are dynamically updated according to the evaluation results to ensure that the algorithm can efficiently search for the optimal solution. By repeating the above evaluation and update process until the maximum number of iterations is reached, the integrity and convergence of the training process are guaranteed. In summary, the optimized technical solutions provided in steps T21 to T24 combine the Legendre polynomial with the neural network and introduce the improved rat swarm algorithm for training. It is an innovative modeling and optimization method that can effectively solve the deficiencies of traditional methods in the effectiveness evaluation of complex systems, realize the efficient modeling and optimization of the equipment support system effectiveness evaluation, and improve the evaluation accuracy and model performance.

[0117] As Figure 6 shown, Figure 6 it shows a specific implementation manner of optimizing the LKAN neural network using the improved rat swarm algorithm. In Figure 6 it, the improved rat swarm algorithm is used to adjust the LKAN neural network parameters. The position of the rat corresponds to a combination of the LKAN neural network parameters, and the improved rat swarm algorithm optimizes the LKAN neural network parameter optimization process.

[0118] As Figure 7 shown, Figure 7 it shows the effectiveness evaluation process of the equipment support system optimized based on the LKAN neural network. The actual effect exerted when the equipment support system completes the equipment support task under the existing equipment support capabilities is comprehensively reflected by the effectiveness evaluation, which reflects the effective degree of the equipment support activities in completing the specified tasks. By introducing a new parameter adjustment mechanism and improving the dynamic balance in the search process of the rat swarm algorithm, the search efficiency and global optimization ability of the algorithm are improved.

[0119] In Figure 7 the effectiveness evaluation process of the equipment support system optimized based on the LKAN neural network shown, it at least includes the following steps:

[0120] Based on the influence of modern warfare by high and new technologies and the development trend of the equipment support system, on the basis of investigations in relevant troops and expert opinions, comprehensively considering the main factors in the operation of the equipment support system, an effectiveness evaluation index system for the equipment support system is established.

[0121] On the basis of consulting relevant materials and literature, determine the data samples for simulation experiments through relevant investigations.

[0122] To make the effectiveness evaluation result of the equipment support system more in line with the actual situation and be able to truly reflect the effectiveness of the equipment support system, it is necessary to perform normalization processing on the data samples. The data after normalization processing can eliminate the influence of the index unit and its numerical order of magnitude, reduce the network prediction error, and can also accelerate the network training speed.

[0123] Determine the LKAN network structure. Exemplarily, the number of input layers of the LKAN neural network is the same as the number of effectiveness evaluation indicators of the equipment support system, and there is only 1 evaluation value of the equipment support system effectiveness in the output layer. Train the LKAN neural network through the improved rat swarm algorithm and training sample data.

[0124] The embodiment of the present application also provides an alternative technical solution for the effectiveness evaluation process of the equipment support system optimized based on the LKAN neural network. Any step in this alternative technical solution can be used as any specific implementation manner of the embodiment of the present application. For example, the steps of index system construction and data preprocessing can be applied to the technical solutions in steps S1 to S3, and the determination of the neural network structure can also be applied to any technical solution in steps S1 to S3. The specific steps of this alternative technical solution are as follows:

[0125] Index system construction and data preprocessing: Construct an effectiveness evaluation index system for the equipment support system, determine the data samples for simulation experiments through investigations, and perform normalization processing on the data samples to eliminate the influence of the index unit and its numerical order of magnitude, reduce the network prediction error, and accelerate the network training speed.

[0126] Determination of the neural network structure: Determine the LKAN neural network structure. The number of input layer nodes is the same as the number of effectiveness evaluation indicators of the equipment support system, and the output layer is the evaluation value of the equipment support system effectiveness. Determine the number of hidden layer nodes according to the empirical formula. The empirical formula can be n = m + k + α, where n is the number of hidden layers, m is the number of input layer nodes, k is the number of output layer nodes, and α is a constant with a value range of [1, 10].

[0127] Coding operation: Perform the coding operation in the algorithm using binary coding, represent each value size of the radial basis function center, width, and output weight of the LKAN neural network with multi-bit binary strings respectively, and determine the individual size, that is, the length of the binary coding string, according to the number of input layers, output layers, and hidden layers.

[0128] Population initialization and fitness calculation: Set the population size and the maximum number of iterations, randomly initialize the population, and solve the fitness value of the current individual according to the fitness function.

[0129] In this step, the fitness function can adopt where: x' represents the sample data value after normalization processing, x represents the original sample data value, and x max x min respectively represent the maximum and minimum values of the same index that appear in the sample data.

[0130] Selection operation: Perform a sorting operation on the individuals of the previous generation and the offspring individuals based on the individual fitness values, and select the individuals with large fitness values to form a new generation of population.

[0131] Judge whether the current iteration number is equal to the maximum iteration number. If they are equal, output the best individual of the population and use it as the parameters of the LKAN neural network structure. If they are not equal, continue to perform the sorting operation.

[0132] Crossover and mutation operations: Perform a complete crossover operation on the individuals of the previous generation population and calculate the fitness values of the alternative offspring individuals.

[0133] Perform pre-mutation on the individuals of the previous generation population and the alternative offspring individuals, and then calculate the fitness values of the pre-mutated individuals. If the fitness value of the pre-mutated individual is greater than the fitness value of the current individual, mutation occurs; otherwise, no mutation operation is performed, and the steps of the fitness function are skipped and continued until a mutation operation occurs.

[0134] Result output: Output the best individual of the population and use it as the parameters of the LKAN neural network structure, and apply it to the effectiveness evaluation of the equipment support system.

[0135] The alternative technical solution provided by the embodiments of the present application improves the evaluation accuracy of the equipment support system, enables the neural network to more efficiently approximate complex functional relationships, improves the prediction accuracy of the model, enhances the global search ability, introduces random perturbations and dynamically adjusts parameters to avoid the algorithm falling into local optima, enhances the global search ability of the algorithm, improves the optimization performance, and enables the algorithm to find the global optimal solution within fewer iterations and improve the convergence speed by reasonably setting parameters and optimization strategies. Compared with the defect that the effectiveness evaluation method of the equipment support system in the related technology relies on expert experience and has strong subjective factors, the embodiments of the present application automatically adjust the network parameters through an optimized algorithm, reduce the interference of human factors, ensure that the evaluation results are more objective and accurate, reduce unnecessary calculations, improve the calculation efficiency, and reduce the calculation cost. At the same time, the improved algorithm can find the optimal solution faster, further improving the evaluation efficiency, and also enhancing the global search ability and the ability to adapt to complex problems of the algorithm, providing an efficient and accurate method for the evaluation of the effectiveness of the equipment support system.

[0136] As Figure 8 shown, based on the LKAN neural network optimization method and the equipment support system evaluation method based on the rat swarm algorithm, those skilled in the art also provide corresponding electronic devices and storage media:

[0137] The electronic device includes one or more processors 51 and a memory 52. In the figure, one processor 51 is taken as an example.

[0138] The controller may further include: an input device 53 and an output device 54.

[0139] The processor 51, the memory 52, the input device 53, and the output device 54 may be connected through a bus or other means. In the figure, connection through a bus is taken as an example.

[0140] The processor 51 may be a central processing unit (CPU for short), and the processor 51 may also be other general-purpose processors, digital signal processors (DSP for short), application-specific integrated circuits (ASIC for short), field-programmable gate arrays (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. chips, or a combination of the above types of chips. The general-purpose processor may be a microprocessor or any conventional processor.

[0141] The memory 52, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the control method in the embodiments of the present application. The processor 51 executes various functional applications and data processing of the server by running the non-transitory software programs, instructions, and modules stored in the memory 52, that is, to implement the method in the above method embodiments.

[0142] The memory 52 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the processing device of the server operation, etc. In addition, the memory 52 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory 52 may optionally include a memory remotely provided relative to the processor 51, and these remote memories can be connected to the network connection device through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0143] The input device 53 can receive input digital or character information, and generate key signal inputs related to the user settings and function controls of the processing device of the server. The output device 54 may include a display device such as a display screen.

[0144] One or more modules are stored in the memory 52, and when executed by one or more processors 51, implement the method shown in any specific implementation manner of the embodiments of the present application.

[0145] Those skilled in the art can understand that to implement all or part of the processes in the above method embodiments, it can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium, and when the program is executed, it may include the processes in the above method embodiments. Among them, the storage medium can be a magnetic disk, an optical disc, a read-only memory (ROM), a random access memory (RAM), a flash memory (FM), a hard disk drive (HDD), or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memories.

[0146] In the description of the specification of the embodiments of the present application, the descriptions referring to the terms "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in a suitable manner in any one or more embodiments or examples.

[0147] In addition, the technical solutions between various embodiments of the present application can be combined with each other, but it must be based on the fact that those skilled in the art can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the embodiments of the present application.

[0148] All features disclosed in the embodiments of the present application, or steps in all methods or processes disclosed, except for mutually exclusive features and / or steps, can be combined in any way. Any feature disclosed in the specification of the embodiments of the present application, unless specifically stated, can be replaced by other equivalent or features with similar purposes. That is, unless specifically stated, each feature is only an example of a series of equivalent or similar features. Throughout the specification, the same reference numerals indicate the same elements.

[0149] Those skilled in the art can understand that the modules in the devices in the embodiments can be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and in addition, they can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in the specification of the embodiments of the present application (including the corresponding claims, abstract and drawings) and all processes or units of any method or device so disclosed. Unless otherwise explicitly stated, each feature disclosed in the specification of the embodiments of the present application (including the corresponding claims, abstract and drawings) can be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0150] Although the embodiments of the present application are described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the embodiments of the present application, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A LKAN neural network optimization method based on rat colony algorithm, characterized in that: include: Determine an individual candidate solution and a current optimal solution, and update the individual candidate solution by using the current optimal solution; Performing iterative operations, evaluating the current optimal solution through the objective function, adjusting the individual candidate solutions and the current optimal solution according to the relationship between the individual candidate solutions and the search space, generating an improved rat swarm algorithm and returning a global optimal solution; A Legendre polynomial neural network is constructed, and an input layer of the Legendre polynomial neural network is set according to the global optimal solution.

2. The LKAN neural network optimization method based on the rat colony algorithm according to claim 1, characterized in that: The determining of the individual candidate solutions and the current optimal solution, and updating the individual candidate solutions by using the current optimal solution, further comprises: Initialize the rat swarm algorithm, set the current individual position, the current best position, the maximum number of iterations, the exploration parameter and the development parameter in the position update formula; The exploration and development mechanisms of the rat swarm algorithm are balanced by the exploration parameter and the development parameter; Subtract the current individual position from the current best position, take the absolute value of the difference result, and get the next individual position.

3. The LKAN neural network optimization method based on the rat colony algorithm according to claim 1, characterized in that: The iterative operation is performed, the current optimal solution is evaluated by the objective function, the individual candidate solutions and the current optimal solution are adjusted according to the relationship between the individual candidate solutions and the search space, the improved rat swarm algorithm is generated and the global optimal solution is returned, further comprising: Initialize the parameters of the position update formula in the rat swarm algorithm, including the population size, the maximum number of iterations, and the current global optimal solution; The current individual position of the position update formula in the rat swarm algorithm is detected, and if the current individual position is beyond the search space of the rat swarm algorithm, the current individual position is reallocated back to the previous position.

4. The LKAN neural network optimization method based on the rat colony algorithm according to claim 3, characterized in that: In the process of reallocating the current individual position back to the previous position, the individual candidate solutions are evaluated by an objective function, including: Initialize the candidate solution set and the objective function corresponding to the individual candidate solution, calculate the fitness value of the individual candidate solution through the objective function, and update the position of the individual candidate solution; Dynamically adjust the parameters in the position update formula according to the value of the objective function to keep the individual candidate solutions within the preset search space; Repeat the above steps until the number of iterations reaches the preset maximum number of iterations or the objective function meets the convergence condition.

5. The LKAN neural network optimization method based on the rat colony algorithm according to claim 1, characterized in that: The constructing of the Legendre polynomial neural network, setting the input layer of the Legendre polynomial neural network according to the global optimal solution, is specifically as follows: The Legendre polynomial neural network includes an input layer, a hidden layer and an output layer, wherein the input layer is used to receive a global optimal solution, the hidden layer uses Legendre polynomials as basis functions, and the output layer is used to output optimization results.

6. A method for evaluating an equipment support system, characterized in that: The LKAN neural network optimization method based on the rat swarm algorithm according to any one of claims 1 to 5, further comprising: Acquire data samples, and preprocess the data samples to generate preprocessed data; The Legendre polynomial neural network is set so that the number of input layers of the Legendre polynomial neural network corresponds to the number of equipment support system effectiveness evaluation indicators; The improved rat swarm algorithm is used to train the Legendre polynomial neural network, and the trained Legendre neural network is used to evaluate the effectiveness of the equipment support system to obtain the equipment support system effectiveness evaluation value.

7. The equipment support system evaluation method according to claim 6, characterized in that: The acquiring of data samples and preprocessing of the data samples to generate preprocessed data are specifically as follows: The data samples are normalized to generate normalized data, and the normalized data is used to eliminate the influence of the indicator unit and its numerical magnitude on the data samples.

8. The equipment support system evaluation method according to claim 6, characterized in that: The setting of the Legendre polynomial neural network further includes: In the output layer of the Legendre polynomial neural network, Legendre polynomials are used as basis functions; The Legendre polynomial neural network is trained based on the improved rat swarm algorithm, and the individual positions of the position update formula in the rat swarm algorithm are initialized so that the individual positions correspond to the combination of parameters of the Legendre polynomial neural network; Evaluate the individual position of the position update formula according to the objective function, and update the corresponding individual position according to the evaluation result; Repeat the above steps until the number of iterations reaches the maximum number of iterations.

9. An electronic device, characterized in that: The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor executes the method described in any one of claims 1 to 8.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images