Chip defect classification method for optimizing feature selection based on spot-bird optimization algorithm
By optimizing feature selection based on Kingfisher optimization algorithm, the optimal parameters of the support vector machine are found, which solves the problem of low chip defect classification accuracy and achieves a higher classification accuracy.
Patent Information
- Application Number
- CN202411048195.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the prior art, the chip defect classification accuracy is low, making it difficult to select the best feature combination and support vector machine parameters, affecting the classification performance.
The method of optimizing feature selection based on Kingfisher optimization algorithm is adopted. By initializing the population, exploring, developing and symbiosis iteration process, the optimal punishment factor C and kernel function parameter γ of the support vector machine are found to improve the classification accuracy.
It improves the classification accuracy of the support vector machine, avoids local optimization, and is suitable for the classification tasks of chip surface defects, significantly improving the classification accuracy.
Smart Images

Figure CN118916772A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of machine learning and optimization algorithms, and in particular to a chip defect classification method based on feature selection optimization of a kingfisher optimization algorithm. Background Art
[0002] Feature selection is a key step in machine learning, which aims to select the most relevant subset from a large number of features to improve the performance and efficiency of the model. The main goal of feature selection is to reduce the dimensionality of the data, reduce the computational complexity, and improve the generalization ability of the model by removing redundant and irrelevant features. Support vector machine (SVM) is a supervised learning model for classification and regression analysis, which is widely used in pattern recognition, text classification, image recognition and other fields. However, the performance of SVM is highly dependent on its parameter selection, such as the penalty factor C and the kernel function parameter γ. Therefore, how to choose the best feature combination and SVM parameters becomes a key issue affecting classification performance.
[0003] Optimization algorithms, such as genetic algorithms, particle swarm optimization, and sparrow search algorithms, have been used for feature selection and optimization of SVM parameter configuration. The Pied Kingfisher Optimization Algorithm (PKO) is an emerging bionic intelligent optimization algorithm that simulates the predation behavior of Pied Kingfishers in nature and has good global search and local optimization capabilities. Therefore, the Pied Kingfisher Optimization Algorithm is used to optimize feature selection and find the optimal penalty factor C and kernel function parameter γ of the support vector machine for chip surface defect classification, which can improve the classification accuracy. Summary of the invention
[0004] In order to solve the above technical problems, the present invention proposes a chip defect classification method based on feature selection optimization using a Kingfisher optimization algorithm to solve the problem of low classification accuracy in the prior art.
[0005] To achieve the above object, the technical solution adopted by the present invention is:
[0006] The chip defect classification method based on the Kingfisher optimization algorithm to optimize feature selection is characterized by: the specific steps are as follows;
[0007] Step 1: Read the chip surface defect feature data, determine the optimization target is the feature combination, penalty parameter C and kernel function parameter γ that minimizes the classification error of the test set sample of the support vector machine;
[0008] Step 2: Initialize the parameters of the kingfisher optimization algorithm, including the population size and the maximum number of iterations, and randomly generate the initial population so that it is evenly distributed in the search space;
[0009] Step 3: Use the classification error of the test set samples in the support vector machine method as the fitness function of the kingfisher optimization algorithm, and calculate the fitness function value of each kingfisher individual;
[0010] Step 4: If rand()<0.8, the exploration phase begins. The exploration phase is inspired by the perching and hovering behavior of the kingfisher. In the kingfisher algorithm, the location of the population is determined based on the foraging activities of the kingfisher. If rand()>0.5, the perching strategy is executed, otherwise the hovering strategy is executed. If rand()>0.8, the development phase begins. The development phase is inspired by the diving behavior of the kingfisher. If the newly generated solutions are better than the previous solutions, they are replaced and the symbiosis phase is executed. The symbiosis phase is to improve the diversity of solutions by simulating the symbiotic relationship between the kingfisher and the Eurasian otter.
[0011] Step 5: Determine whether the termination condition is met. If the maximum number of iterations is reached, terminate the iteration; otherwise, return to step 4 to continue iterating;
[0012] Step 6: Output the optimal feature combination, penalty parameter C, kernel function parameter γ, and classification accuracy obtained in step 5.
[0013] Further, the initialization of the kingfisher optimization algorithm parameters in step 2 is specifically as follows:
[0014] The position of each kingfisher is initialized as:
[0015] X i,j =LB+(UB-LB)·rand (1)
[0016] Where X i,j represents the position of the i-th individual in the j-th dimension, rand is a random value between 0 and 1, UB and LB are the upper and lower limits of the search range, respectively. The position of each kingfisher represents a candidate solution to the optimization problem.
[0017] Furthermore, in step 3, the classification accuracy of the training set samples in the support vector machine method is used as the fitness function of the kingfisher optimization algorithm. The fitness function is specifically:
[0018] For a support vector X s , its distance to the hyperplane can be expressed as;
[0019]
[0020] This distance is called a "hard interval", which correctly divides all sample points and the samples are outside the interval boundary. At this time, |w T x s +b|=1, so the problem of maximizing the interval can be transformed into the problem of finding Max(1 / ||w||), with the constraint that |w T x i +b|>=1,i=1,...,n. Among them, xi is the position of the sample in the feature space;
[0021] For the convenience of calculation, the optimization problem of the support vector machine to find the optimal hyperplane can be described as follows:
[0022]
[0023] The restriction condition is y i (w T x i +b)≥1,i=1,...,n;
[0024] For each training sample and label (x i ,y i ), set the slack variable ε i , then the problem of nonlinear support vector machine finding the optimal hyperplane can be described as follows:
[0025]
[0026] The restriction condition is changed to y i (w T x i +b)≥1-ε i ,ε i ≥0,i=1,...,n, where C is the penalty factor, which represents the tolerance of the support vector machine to misclassified samples and is a constant greater than 0;
[0027] After the constrained problem of soft margin maximization is transformed into an unconstrained problem using the Lagrangian function and then transformed into an equivalent dual form, the problem is transformed into the following form:
[0028]
[0029] The restriction condition is
[0030] Let X be a low-dimensional space, input variable x i ,x j ∈X, Y is the transformed high-dimensional feature space, then the kernel function is defined as:
[0031]
[0032] At this time, the nonlinear support vector machine optimization problem is transformed into the following form:
[0033]
[0034] The restriction condition is still However, the inner product operation becomes a kernel function operation, which greatly reduces the amount of calculation;
[0035] The mainstream kernel function of nonlinear support vector machine is radial basis kernel function, the formula is as follows:
[0036]
[0037] The smaller γ is, the more specific the classification is, which may lead to overfitting; the larger γ is, the more vague the classification is, which may lead to underfitting. Therefore, it is necessary to correctly define the value of γ to improve the classification performance of the support vector machine.
[0038] Furthermore, in step 3, the classification error of the test set samples is used as the fitness function of the kingfisher optimization algorithm. Specifically, the fitness function value of each kingfisher individual is calculated. The goal of the kingfisher optimization algorithm is to find a set of feature subsets, penalty parameters C and kernel function parameters γ, so that the error of the test set is minimized and the accuracy is maximized. The test set error is defined as:
[0039]
[0040] Among them, T sim represents the predicted label, t test represents the true label.
[0041] Furthermore, the exploration phase described in step 4 is specifically as follows:
[0042] In the exploration phase, the individual position will be adjusted according to the situation to make it tend to the global optimal solution. In this phase, the search individual updates its position according to the foraging behavior of the kingfisher. The update formula is shown in formula (10):
[0043] X i (t+1)=X i (t)+a*T′(X j (t)-X i (t)) (10)
[0044] Among them, X i (t+1) represents the solution of the next iteration, X i (t) represents the solution of the current iteration. The parameter α is solved by 2′randn(1,Dim)-1, where randn represents a random number in the normal distribution and Dim represents the dimension of the solution. If rand()>0.5, the roosting strategy is executed and T is calculated by formula (11):
[0045]
[0046] Otherwise, the hovering strategy is executed and T is calculated by formula (12):
[0047]
[0048] Among them, PKO_Fitness(i) and PKO_Fitness(j) are the fitness values of the i-th and j-th kingfishers respectively. In addition, BF is the flapping factor, which is usually set to 8.
[0049] Furthermore, the development phase described in step 4 is specifically as follows:
[0050] During the development phase, the diving behavior of the kingfisher was simulated for local search, that is, a more detailed search near the optimal solution that has been found. When the kingfisher finds a fish, it will quickly dive into the water to catch it. This behavior is modeled by the following mathematical formula:
[0051] X i (t+1)=X i (t)+HA*o*a*(bX best (t)) (13)
[0052] Among them, α is a control parameter calculated by 2′randn(1,Dim)-1, b=X i (t)+o 2 *randn*X best (t), PKO_Fitness(i) represents the fitness function value of the ith kingfisher, Best_Fitness represents the best fitness value in all iterations, and o&HA represents the hunting ability of the kingfisher.
[0053] Furthermore, the symbiotic stage described in step 4 is specifically as follows:
[0054] If the newly generated solution is closer to the optimal solution than the original solution, then the original solution is replaced and the symbiotic phase is executed. The symbiotic phase is inspired by the symbiotic relationship between otters and kingfishers, which allows the kingfisher to benefit from the hunting behavior of the otter, which is mathematically represented as:
[0055]
[0056] Among them, 2 individuals are randomly selected from the population, and their positions are represented by X. m and X n The predation efficiency of kingfisher is represented by PE. Among them, PE max and PE min Fixed values are 0.5 and 0 respectively.
[0057] Compared with the prior art, the present invention has the following advantages:
[0058] 1. The optimal feature subset and support vector machine parameters were selected to improve the accuracy of support vector machine classification;
[0059] 2. The global search capability of the kingfisher optimization algorithm is strong and can avoid falling into the local optimum;
[0060] 3. Suitable for classification tasks of chip surface defects. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] Figure 1 is a flow chart of the method of the present invention;
[0062] Figure 2 It is a confusion matrix result diagram of the test set data obtained by the method of the present invention;
[0063] Figure 3 It is a comparison chart of the true labels of the test set data obtained by the method of the present invention and the SVM predicted labels. DETAILED DESCRIPTION
[0064] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:
[0065] As a specific embodiment of the present invention, the present invention adopts the following technical solution, wherein the flow chart is as follows Figure 1 As shown:
[0066] A chip defect classification method based on feature selection optimization using the Kingfisher Optimization Algorithm. The Kingfisher Optimization Algorithm simulates the predation behavior of the Kingfisher, updates the positions of population members through the three stages of exploration, development, and symbiosis, and strikes a balance between global search and local optimization to find the optimal solution to the problem. The Kingfisher Optimization Algorithm is used to optimize the feature subset, the penalty parameter C of the support vector machine, and the kernel function parameter γ, and is applied to the classification of chip surface defects. The specific steps include:
[0067] Step 1: Read the chip surface defect feature data, determine the optimization target is the feature combination, penalty parameter C and kernel function parameter γ that minimizes the classification error of the test set sample of the support vector machine;
[0068] Step 2: Initialize the parameters of the kingfisher optimization algorithm, including population size, maximum number of iterations, etc. Randomly generate the initial population so that it is evenly distributed in the search space;
[0069] Step 3: Use the classification error of the test set samples in the support vector machine method as the fitness function of the kingfisher optimization algorithm, and calculate the fitness function value of each kingfisher individual;
[0070] Step 4: If rand()<0.8, the exploration phase begins. The exploration phase is inspired by the perching and hovering behavior of the kingfisher. In the kingfisher algorithm, the location of the population is determined based on the foraging activities of the kingfisher. If rand()>0.5, the perching strategy is executed, otherwise the hovering strategy is executed. If rand()30.8, the exploitation phase begins. The exploitation phase is inspired by the diving behavior of the kingfisher. If the newly generated solutions are better than the previous ones, they are replaced and the symbiosis phase is executed. The symbiosis phase is to improve the diversity of solutions by simulating the symbiotic relationship between the kingfisher and the Eurasian otter.
[0071] Step 5: Determine whether the termination condition is met. If the maximum number of iterations is reached, terminate the iteration; otherwise, return to step 4 to continue iterating;
[0072] Step 6: Output the optimal feature combination, penalty parameter C and kernel function parameter γ obtained in step 5, as well as the classification accuracy. Further, the parameters of the kingfisher optimization algorithm are initialized in step 2. The position of each kingfisher is initialized to:
[0073] X i,j =LB+(UB-LB)·rand
[0074] Where X i,j represents the position of the i-th individual in the j-th dimension, rand is a random value between 0 and 1, UB and LB are the upper and lower limits of the search range, respectively. The position of each kingfisher represents a candidate solution to the optimization problem.
[0075] Furthermore, the classification error of the test set samples described in step 3 is used as the fitness function of the kingfisher optimization algorithm, and the fitness function value of each kingfisher individual is calculated. The goal of the kingfisher optimization algorithm is to find a set of feature subsets, penalty parameters C and kernel function parameters γ, so that the error of the test set is minimized and the accuracy is maximized. The test set error can be defined as:
[0076]
[0077] Among them, T sim represents the predicted label, t test represents the true label.
[0078] Further, the exploration phase described in step 4. The exploration phase will adjust the individual position according to the situation to make it tend to the global optimal solution. In this phase, the search individual updates the position according to the foraging behavior of the kingfisher. The update formula is:
[0079] X i (t+1)=X i (t)+a*T′(X j (t)-X i(t))
[0080] Among them, X i (t+1) represents the solution of the next iteration, X i (t) represents the solution of the current iteration. The parameter α is solved by 2′randn(1,Dim)-1, where randn represents a random number in a normal distribution and Dim represents the dimension of the solution. If rand()>0.5, the roosting strategy is executed and T is calculated by the following formula:
[0081]
[0082] Otherwise, the hover strategy is executed and T is calculated as follows:
[0083]
[0084] Among them, PKO_Fitness(i) and PKO_Fitness(j) are the fitness values of the i-th and j-th kingfishers respectively. In addition, BF is the flapping factor, which is usually set to 8.
[0085] Further, the development phase described in step 4. The development phase simulates the diving behavior of the kingfisher, which is mainly used for local search, that is, a more detailed search near the better solution that has been found to improve the accuracy of the solution. When the kingfisher finds a fish, it will quickly dive into the water to catch the fish. This behavior can be modeled by the following mathematical formula:
[0086] X i (t+1)=X i (t)+HA*o*a*(bX best (t))
[0087] Among them, α is a control parameter calculated by 2′randn(1,Dim)-1, b=X i (t)+o 2 *randn*X best (t), PKO_Fitness(i) represents the fitness function value of the ith kingfisher, Best_Fitness represents the best fitness value in all iterations. o&HA represents the hunting ability of the kingfisher.
[0088] Further, the symbiotic phase described in step 4. If the newly generated solution is closer to the optimal solution than the original solution, then replace the original solution and execute the symbiotic phase. The symbiotic phase is inspired by the symbiotic relationship between otters and kingfishers, which allows kingfishers to benefit from the hunting behavior of otters. This behavior is mathematically expressed as:
[0089]
[0090] Among them, 2 individuals are randomly selected from the population, and their positions are represented by X. m and X n The predation efficiency of the kingfisher is represented by PE. Among them, PE max and PE min Fixed values are 0.5 and 0 respectively. DETAILED DESCRIPTION
[0092] Step 1: Read the chip surface defect feature data, determine the optimization target is the feature combination, penalty parameter C and kernel function parameter γ that minimizes the classification error of the test set sample of the support vector machine;
[0093] Step 2: Initialize the parameters of the kingfisher optimization algorithm, including population size, maximum number of iterations, etc. Randomly generate the initial population so that it is evenly distributed in the search space; the position of each kingfisher is initialized to:
[0094] X i,j =LB+(UB-LB)·rand
[0095] Where X i,j represents the position of the i-th individual in the j-th dimension, rand is a random value between 0 and 1, UB and LB are the upper and lower limits of the search range, respectively. The position of each kingfisher represents a candidate solution to the optimization problem.
[0096] Step 3: Use the classification error of the test set samples in the support vector machine method as the fitness function of the kingfisher optimization algorithm, and calculate the fitness function value of each kingfisher individual; the goal of the kingfisher optimization algorithm is to find a set of feature subsets, penalty parameters C and kernel function parameters γ, so that the error of the test set is minimized and the accuracy is maximized. The test set error can be defined as:
[0097]
[0098] Among them, T sim represents the predicted label, t test represents the true label.
[0099] Step 4: If rand()<0.8, the exploration phase begins. The exploration phase is inspired by the roosting and hovering behavior of the kingfisher. In the kingfisher algorithm, the location of the population is determined based on the foraging activities of the kingfisher. The exploration phase adjusts the individual positions according to the situation to make them tend to the global optimal solution. If rand()>0.5, the roosting strategy is executed, otherwise the hovering strategy is executed. In this phase, the search individual updates its position based on the foraging behavior of the kingfisher, and the update formula is:
[0100] X i(t+1)=X i (t)+a*T′(X j (t)-X i (t))
[0101] Among them, X i (t+1) represents the solution of the next iteration, X i (t) represents the solution of the current iteration. The parameter α is solved by 2′randn(1,Dim)-1, where randn represents a random number in a normal distribution and Dim represents the dimension of the solution. If rand()>0.5, the roosting strategy is executed and T is calculated by the following formula:
[0102]
[0103] Otherwise, the hover strategy is executed and T is calculated as follows:
[0104]
[0105] Among them, PKO_Fitness(i) and PKO_Fitness(j) are the fitness values of the ith and jth kingfisher, respectively. In addition, BF is the flapping factor, which is usually set to 8.
[0106] If rand() is 30.8, the development phase begins. The development phase is inspired by the diving behavior of the kingfisher. The development phase simulates the diving behavior of the kingfisher and is mainly used for local search, that is, to conduct a more refined search near the better solution that has been found to improve the accuracy of the solution. When the kingfisher finds a fish, it will quickly dive into the water to catch the fish. This behavior can be modeled by the following mathematical formula:
[0107] X i (t+1)=X i (t)+HA*o*a*(bX best (t))
[0108] Among them, α is a control parameter calculated by 2′randn(1,Dim)-1, b=X i (t)+o 2 *randn*X best (t), PKO_Fitness(i) represents the fitness function value of the ith kingfisher, Best_Fitness represents the best fitness value in all iterations. o&HA represents the hunting ability of the kingfisher.
[0109] If the newly generated solutions are better than the previous ones, they are replaced and the symbiosis phase is executed. The symbiosis phase is to improve the diversity of solutions by simulating the symbiotic relationship between the kingfisher and the Eurasian otter. The symbiosis phase is inspired by the symbiotic relationship between the otter and the kingfisher, which allows the kingfisher to benefit from the otter's hunting behavior. This behavior is mathematically represented as:
[0110]
[0111] Among them, 2 individuals are randomly selected from the population, and their positions are represented by X. m and X n The predation efficiency of the kingfisher is represented by PE. Among them, PE max and PE min Fixed values are 0.5 and 0 respectively.
[0112] Step 5: Determine whether the termination condition is met. If the maximum number of iterations is reached, terminate the iteration; otherwise, return to step 4 to continue iterating;
[0113] Step 6: Output the optimal feature combination, penalty parameter C, kernel function parameter γ, and classification accuracy obtained in step 5.
[0114] The confusion matrix result of the test set data of the method of the present invention is shown in the figure Figure 2 As shown in the figure, the comparison of the actual label of the test set data and the SVM predicted label is as follows Figure 3 The experimental results show that the present invention can effectively classify chip surface defects, and the method of optimizing feature selection based on the Kingfisher optimization algorithm has a higher test classification accuracy.
[0115] The present invention proposes a chip defect classification method based on Kingfisher optimization algorithm to optimize feature selection, and utilizes the global search capability of Kingfisher optimization algorithm to optimize feature subsets, penalty parameter C of support vector machine and kernel function parameter γ, so as to effectively improve the classification accuracy of support vector machine.
[0116] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.
Claims
1. A chip defect classification method based on feature selection optimized by Kingfisher optimization algorithm, characterized by: The specific steps are as follows; Step 1: Read the chip surface defect feature data, determine the optimization target is the feature combination, penalty parameter C and kernel function parameter γ that minimizes the classification error of the test set sample of the support vector machine; Step 2: Initialize the parameters of the kingfisher optimization algorithm, including the population size and the maximum number of iterations, and randomly generate the initial population so that it is evenly distributed in the search space; Step 3: Use the classification error of the test set samples in the support vector machine method as the fitness function of the kingfisher optimization algorithm, and calculate the fitness function value of each kingfisher individual; Step 4: If rand()<0.8, the exploration phase begins. The exploration phase is inspired by the perching and hovering behavior of the kingfisher. In the kingfisher algorithm, the location of the population is determined based on the foraging activities of the kingfisher. If rand()>0.5, the perching strategy is executed, otherwise the hovering strategy is executed. If rand()>0.8, the development phase begins. The development phase is inspired by the diving behavior of the kingfisher. If the newly generated solutions are better than the previous solutions, they are replaced and the symbiosis phase is executed. The symbiosis phase is to improve the diversity of solutions by simulating the symbiotic relationship between the kingfisher and the Eurasian otter. Step 5: Determine whether the termination condition is met. If the maximum number of iterations is reached, terminate the iteration; otherwise, return to step 4 to continue iterating; Step 6: Output the optimal feature combination, penalty parameter C, kernel function parameter γ, and classification accuracy obtained in step 5.
2. The chip defect classification method based on the Kingfisher optimization algorithm optimized feature selection according to claim 1 is characterized in that: The initialization of the kingfisher optimization algorithm parameters in step 2 is specifically as follows: The position of each kingfisher is initialized as: X i,j =LB+(UB-LB)·rand (1) Where X i,j represents the position of the i-th individual in the j-th dimension, rand is a random value between 0 and 1, UB and LB are the upper and lower limits of the search range, respectively. The position of each kingfisher represents a candidate solution to the optimization problem.
3. The chip defect classification method based on the Kingfisher optimization algorithm optimized feature selection according to claim 1 is characterized in that: In step 3, the classification accuracy of the training set samples in the support vector machine method is used as the fitness function of the kingfisher optimization algorithm. The fitness function is specifically: For a support vector X s , its distance to the hyperplane can be expressed as; This distance is called "hard interval", which correctly divides all sample points and the samples are outside the interval boundary. At this time, |w T x s +b|=1, so the problem of maximizing the interval can be transformed into the problem of finding Max(1 / ||w||), with the constraint w T x i +b|>=1,i=1,...,n. Among them, x i is the position of the sample in the feature space; For the convenience of calculation, the optimization problem of the support vector machine to find the optimal hyperplane can be described as follows: The restriction condition is y i (w T x i +b)≥1,i=1,...,n; For each training sample and label (x i ,y i ), set the slack variable ε i , then the problem of nonlinear support vector machine finding the optimal hyperplane can be described as follows: The restriction condition is changed to y i (w T x i +b)≥1-ε i ,ε i ≥0,i=1,...,n, where C is the penalty factor, which represents the tolerance of the support vector machine to misclassified samples and is a constant greater than 0; After the constrained problem of soft margin maximization is transformed into an unconstrained problem using the Lagrangian function and then transformed into an equivalent dual form, the problem is transformed into the following form: The restriction condition is Let X be a low-dimensional space, input variable x i ,x j ∈X, Y is the transformed high-dimensional feature space, then the kernel function is defined as: At this time, the nonlinear support vector machine optimization problem is transformed into the following form: The restriction condition is still However, the inner product operation becomes a kernel function operation, which greatly reduces the amount of calculation; The mainstream kernel function of nonlinear support vector machine is radial basis kernel function, the formula is as follows: The smaller γ is, the more specific the classification is, which may lead to overfitting; the larger γ is, the more vague the classification is, which may lead to underfitting. Therefore, it is necessary to correctly define the value of γ to improve the classification performance of the support vector machine.
4. The chip defect classification method based on the Kingfisher optimization algorithm optimized feature selection according to claim 1 is characterized in that: Step 3: The classification error of the test set samples is used as the fitness function of the kingfisher optimization algorithm. Specifically, the fitness function value of each kingfisher individual is calculated. The goal of the kingfisher optimization algorithm is to find a set of feature subsets, penalty parameters C and kernel function parameters γ, so that the error of the test set is minimized and the accuracy is maximized. The test set error is defined as: Among them, T sim represents the predicted label, t test represents the true label.
5. The chip defect classification method based on the Kingfisher optimization algorithm optimized feature selection according to claim 1 is characterized in that: The exploration phase described in step 4 is specifically as follows: In the exploration phase, the individual position will be adjusted according to the situation to make it tend to the global optimal solution. In this phase, the search individual updates its position according to the foraging behavior of the kingfisher. The update formula is shown in formula (10): X i (t+1)=X i (t)+a*T′(X j (t)-X i (t)) (10) Among them, X i (t+1) represents the solution of the next iteration, X i (t) represents the solution of the current iteration. The parameter α is solved by 2′randn(1,Dim)-1, where randn represents a random number in the normal distribution and Dim represents the dimension of the solution. If rand()>0.5, the roosting strategy is executed and T is calculated by formula (11): Otherwise, the hovering strategy is executed and T is calculated by formula (12): Among them, PKO_Fitness(i) and PKO_Fitness(j) are the fitness values of the i-th and j-th kingfishers respectively. In addition, BF is the flapping factor, which is usually set to 8.
6. The chip defect classification method based on the Kingfisher optimization algorithm optimized feature selection according to claim 1 is characterized in that: The development phases described in step 4 are as follows: During the development phase, the diving behavior of the kingfisher was simulated for local search, that is, a more detailed search near the optimal solution that has been found. When the kingfisher finds a fish, it will quickly dive into the water to catch it. This behavior is modeled by the following mathematical formula: X i (t+1)=X i (t)+HA*o*a*(b-X best (t)) (13) Among them, α is a control parameter calculated by 2′randn(1,Dim)-1, b=X i (t)+o 2 *randn*X best (t), PKO_Fitness(i) represents the fitness function value of the ith kingfisher, Best_Fitness represents the best fitness value in all iterations, and o&HA represents the hunting ability of the kingfisher.
7. The chip defect classification method based on feature selection optimization of Kingfisher optimization algorithm according to claim 1 is characterized in that: The symbiotic stage described in step 4 is specifically: If the newly generated solution is closer to the optimal solution than the original solution, then the original solution is replaced and the symbiotic phase is executed. The symbiotic phase is inspired by the symbiotic relationship between otters and kingfishers, which allows the kingfisher to benefit from the hunting behavior of the otter, which is mathematically represented as: Among them, 2 individuals are randomly selected from the population, and their positions are represented by X. m and X n The predation efficiency of kingfisher is represented by PE. Among them, PE max and PE min Fixed values are 0.5 and 0 respectively.
Citation Information
Patent Citations
Method for constructing prediction model based on orthogonal multi-population sine and cosine algorithm
CN110222751A
Electric energy quality disturbance source judgment method based on bird flock algorithm and SVM
CN110648088A
Method for constructing prediction model based on hunger game search algorithm
CN110751257A
Lithium battery surface defect identification method based on PCA-IAOA-SVM
CN116010870A
Cited By
Rolling bearing vibration signal prediction method based on improved green bird optimization network model
CN120995219A
Laser spot detection method based on secretary bird optimization algorithm
CN121033091A
