Characteristic selection method and system based on improved hornink optimization algorithm

Abstract

The invention discloses a feature selection method and a feature selection system based on an improved Kerr-Rink optimization algorithm, and relates to the technical field of data feature selection, and the method comprises the following steps: updating the position of an individual by using an alpha-melanotropin rate (MSH), a BLS based on biological geography learning strategy, a hierarchical structure (HS), a Kerr-Rink optimization algorithm (HLOA), general reverse learning (OBL) and an immediate following RF; the method comprises the following steps of: constructing an improved angle exid optimization algorithm bROBEHLOA; according to a medical data set, an algorithm population is initialized, the fitness value of each individual of the population is obtained by using a proximity algorithm KNN model, an optimal feature subset is searched by using a bROBEHLOA algorithm, and the optimal feature subset is output when the maximum number of iterations is reached. According to the invention, key feature omission caused by the algorithm falling into a local optimal solution can be avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of data feature selection, and in particular to a feature selection method and system based on an improved horned lizard optimization algorithm. Background Art

[0002] Medical data generally has the characteristics of high dimensionality, multi-source heterogeneity and dynamic evolution. For example, a single whole genome sequencing can generate more than 30,000 feature variables, and the real-time monitoring equipment in the intensive care unit generates new time series data every second. However, the features of real clinical value in these massive data are often submerged in noise and redundant information, resulting in reduced performance of machine learning models, a surge in computing costs, and weakened clinical interpretability. How to efficiently extract key features from complex medical data has become one of the core bottlenecks restricting the development of medical artificial intelligence.

[0003] In the prior art, feature selection technology is the key means to solve this problem. Its core goal is to identify the most discriminative feature subsets for disease prediction or diagnosis through a systematic screening process. In the medical field, feature selection is not only related to the improvement of model accuracy, but also directly affects the credibility of clinical decision-making. However, traditional feature selection methods often have poor effects when dealing with high-dimensional medical data. The fusion of metaheuristic algorithms and wrapper frameworks has created a new idea for wrapper methods. Traditional wrapper methods can improve screening quality by recursively evaluating feature subsets, but their exhaustive search strategy causes the computational complexity to increase exponentially. The heuristic algorithm, due to its efficient global search capability and intelligent optimization mechanism, can not only retain the core advantage of the wrapper method in capturing feature correlation effects, but also significantly reduce the computational cost by dynamically adjusting the search space. However, traditional metaheuristic algorithms still have obvious limitations when dealing with medical scenarios: they are prone to falling into local optimal solutions, resulting in the omission of key features, and insufficient convergence accuracy affects clinical discrimination. The Horned Lizard Optimization Algorithm (HLOA) is a swarm intelligence optimization algorithm that works well in dealing with feature selection problems. However, it also encounters the above problems when facing high-dimensional data sets, such as falling into local optimality, missing key features, and poor convergence accuracy.

[0004] In summary, how to prevent the algorithm from falling into a local optimal solution and causing the omission of key features is an important issue that needs to be solved urgently. Summary of the invention

[0005] The embodiment of the present invention provides a feature selection method and system based on an improved horned lizard optimization algorithm, which can solve the problem in the prior art that the algorithm cannot avoid falling into a local optimal solution, resulting in omission of key features.

[0006] An embodiment of the present invention provides a feature selection method based on an improved horned lizard optimization algorithm, including the following steps: Construct an improved horned lizard optimization algorithm bROBEHLOA; wherein, the improvement of the horned lizard optimization algorithm includes the following steps: use BLS to obtain the sample vector index of each individual, and update the individual position according to the sample vector index; sort the updated individual positions, obtain the three optimal individuals in the sorting, and use the hierarchical structure HS to update the three optimal individuals; use the general opposition-based learning OBL to increase the search range of the algorithm; according to the increased search range, use the random following RF to update the individuals; Perform feature selection on the obtained medical data set, and initialize each individual representing the feature subset in the algorithm population; use the trained K-nearest neighbor algorithm KNN model to obtain the classification result of the feature subset, and use the classification accuracy as the fitness value of each individual; use the bROBEHLOA algorithm to find the feature subset with the optimal fitness value from the initialized population, and output the optimal feature subset when the maximum number of iterations is reached.

[0007] Further, the training steps of the K-nearest neighbor algorithm KNN model include: restricting the historical medical data set to [-1, 1] and dividing the data set into a test set and a training set through ten-fold cross-validation; Use the training set to train the KNN model to obtain the trained KNN model.

[0008] Further, the specific steps of using BLS to obtain the sample vector index of each individual and updating the individual position according to the sample vector index include: Obtain the migration rate and emigration rate of each individual: ; Among them, represents the migration rate; represents the emigration rate; represents the total number of individuals; Generate a sample vector for each individual according to the migration rate and emigration rate; Update the positions of all individuals according to the sample vector: ; ; Among them, represents the position of the current individual under BLS; represents the position of the i th individual updated by the sample vector; is a dim-dimensional randomly generated [0, 1] vector; represents the individual sorted according to the fitness value.

[0009] Furthermore, the steps of sorting the updated individual positions, obtaining the top three individuals in the sorting, and updating the top three individuals using the hierarchical structure HS specifically include: Sort all individuals and select the top three individuals; Use HS to update the top three individuals; ; ; ; ; Among them, The individual ranked first after sorting is named individual A; Is the global optimum under the current iteration; And Are two randomly selected individuals and ; Represents the current iteration number; Represents the maximum number of iterations; Is the individual ranked second after sorting named B The j th dimension value of individual; Represents a randomly selected dimension; Is the individual ranked third after sorting named C The j th dimension value of individual; And Respectively represent the j th dimension values of two randomly selected individuals, and Respectively represent the k th dimension values of individuals A, B, and C; Represents a random number with a value range in [0,1].

[0010] Furthermore, the steps of using the general opposition-based learning OBL to increase the search range specifically include: Use OBL to obtain the opposite point of the individual and obtain the fitness value of the opposite point; According to the fitness value of the opposite point, use the greedy idea to determine whether to retain the opposite point; ; Among them, Represents the opposite point, Represents the current position of the individual, Represents the upper limit of the solution space, Represents the lower limit of the solution space.

[0011] Further, the classification result of the feature subset is obtained by using the trained KNN model of the proximity algorithm, and the classification accuracy is used as the fitness value of each individual. The specific steps are as follows: Convert each individual vector in the training set into a binary vector through the transformation function T function; according to the binary vector, use the trained KNN model to obtain the classification accuracy as the fitness value of each individual.

[0012] An embodiment of the present invention provides a feature selection system based on an improved horned lizard optimization algorithm, including: An algorithm improvement module for constructing an improved horned lizard optimization algorithm bROBEHLOA; wherein, the improvement of the horned lizard optimization algorithm by this algorithm includes the following steps: using BLS to obtain the sample vector index of each individual, and updating the individual position according to the sample vector index; sorting the updated individual positions, obtaining the three optimal individuals in the sorting, and using the hierarchical structure HS to update the three optimal individuals; using the general opposition-based learning OBL to increase the search range of the algorithm; according to the increased search range, using the random following RF to update the individuals; A feature selection module for performing feature selection on the obtained medical data set, initializing each individual representing the feature subset in the algorithm population; obtaining the classification result of the feature subset by using the trained KNN model of the proximity algorithm, and using the classification accuracy as the fitness value of each individual; using the bROBEHLOA algorithm to find the feature subset with the optimal fitness value from the initialized population, and outputting the optimal feature subset when the maximum number of iterations is reached.

[0013] An embodiment of the present invention provides a feature selection method and system based on an improved horned lizard optimization algorithm. Compared with the prior art, its beneficial effects are as follows: The improved horned lizard optimization algorithm bROBEHLOA includes the following steps: using BLS to obtain the sample vector index of each individual, and updating the individual position according to the sample vector index; sorting the updated individual positions, obtaining the three optimal individuals in the sorting, and using the hierarchical structure HS to update the three optimal individuals; using the general opposition-based learning OBL to increase the search range of the algorithm; according to the increased search range, using the random following RF to update the individuals; when using the bROBEHLOA algorithm to find the optimal feature subset, the random following RF and the general opposition-based learning OBL can avoid the algorithm from falling into local optimum and missing key features, and can obtain the optimal feature subset in the final feature selection result. Description of the Drawings

[0014] Figure 1 It is a flowchart of performing feature selection using bROBEHLOA on a public data set provided by an embodiment of the present invention; Figure 2The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of blood; Figure 3 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of lung cancer three-classification; Figure 4 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of Parkinson's; Figure 5 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of wine; Figure 6 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of ionosphere; Figure 7 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of breast cancer cells; Figure 8 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of Cleveland heart disease; Figure 9 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of zoo; Figure 10 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of heart disease; Figure 11 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a low-dimensional public dataset of hepatitis; Figure 12 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on a high-dimensional public dataset of prostate tumors; Figure 13 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of brain tumor 1; Figure 14 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of brain tumor 2; Figure 15 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of small round cell malignant tumor; Figure 16 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of central nervous system; Figure 17 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of diffuse large B-cell lymphoma; Figure 18 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of leukemia; Figure 19 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of leukemia 1; Figure 20 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of leukemia 2; Figure 21 The convergence curve result graph of bROBEHLOA of a feature selection method based on an improved horned lizard optimization algorithm provided by an embodiment of the present invention and each algorithm on the high-dimensional public dataset of lung cancer. Detailed implementation manners

[0015] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following detailed description of the specific embodiments of the present invention will be provided in conjunction with the accompanying drawings. Many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0016] See Figure 1 , an embodiment of the present invention provides a feature selection method based on an improved horned lizard optimization algorithm, including the following steps: Step 1: Construct an improved horned lizard optimization algorithm bROBEHLOA; wherein, the improvement of the horned lizard optimization algorithm by this algorithm includes the following steps: Use BLS to obtain the sample vector index of each individual, and update the individual position according to the sample vector index; Sort the updated individual positions, obtain the three optimal individuals in the sorting, and use the hierarchical structure HS to update the three optimal individuals; Use opposition-based learning (OBL) to increase the search range of the algorithm; According to the increased search range, use random following (RF) to update the individuals.

[0017] Step 2: Perform feature selection on the obtained medical data set, and initialize each individual representing the feature subset in the algorithm population; Use the trained K-nearest neighbor (KNN) model to obtain the classification result of the feature subset, and use the classification accuracy as the fitness value of each individual; Use the bROBEHLOA algorithm to find the feature subset with the optimal fitness value from the initialized population, and output the optimal feature subset when the maximum number of iterations is reached.

[0018] To solve the problem that the HLOA is prone to falling into local optima, the present invention introduces three strategies: random following (RF), opposition-based learning (OBL), and biogeography-based learning strategy (BLS) to strengthen the exploration ability of the HLOA and thus reduce the probability of falling into local optima. In addition, a hierarchical structure (HS) mechanism is introduced to balance the exploration and exploitation stages and accelerate the convergence speed. At the same time, in order to enable the improved HLOA to handle discrete feature selection problems, it is converted into a binary algorithm through a transformation function. The present invention names the improved algorithm bROBEHLOA.

[0019] As Figure 1 shown, this technology mainly includes the following steps: Step 1: Read the medical data set.

[0020] Step 2: Limit the data to [-1, 1] and divide the dataset into a test set and a training set through ten-fold cross-validation.

[0021] Step 3: Initialize the dimension dim, the maximum number of iterations Maxiter, and the optimal vector set Xbest within the parameter range.

[0022] Step 4: Initialize the algorithm population.

[0023] Step 5: Train the training set through the KNN model and calculate the fitness value of each individual.

[0024] Calculating the individual fitness value with the help of the KNN model requires converting each individual vector into a binary vector through a transformation function and then calculating its fitness value. The specific calculation is as follows: .

[0025] .

[0026] .

[0027] .

[0028] Among them, T The function is the transformation function formula, is the i dimensional value of the j th population individual; is the classification accuracy rate; is the individual fitness value; and are set to 0.95 and 0.05; represents the number of features selected by this subset; represents the total number of features.

[0029] Step 6: Use bROBEHLOA to find the optimal feature subset.

[0030] Step 7: Determine whether the current number of iterations exceeds the maximum number of iterations. If not, repeat Step 5 and Step 6.

[0031] Step 8: Output the optimal feature subset.

[0032] Among them, Step 6 is specifically as follows: (1) Generate the α-melanocyte-stimulating hormone rate (MSH).

[0033] The MSH level affects the change in the skin brightness of the horned lizard. Its main purpose is to enhance the exploration ability near the current solution. Its mathematical model is as follows: .

[0034] Among them, and represent the maximum fitness value and the minimum fitness value of the current iteration, respectively; is the k th individual's fitness value. When is less than 0.3, it indicates that the individual has deviated from the development area of the optimal solution and needs to be updated using the formula as follows: .

[0035] Among them, is the i th individual's position in the solution space, is the current optimal solution, and represent two randomly selected individual numbers and they are not equal, is randomly assigned 0 or 1 with equal probability.

[0036] (2) Generate sample vector indices through BLS and update the individual positions with the help of sample vectors.

[0037] BLS updates individuals by combining the generated sample vectors with historical iteration information, which can strengthen the development of the algorithm near the optimal solutions ignored in the early stage and thus reduce the possibility of falling into local optima. The specific calculation model is as follows: First, calculate the migration rate and the removal rate for each individual, and the formulas are as follows: .

[0038] Among them, represents the migration rate; represents the removal rate; represents the total number of individuals. Based on the obtained migration rate and removal rate, generate sample vectors for each individual and generate sample vector indices. The generation rule of this index needs to meet the following conditions: The best individual learns from its own historical best position; the worst individual learns from the historical best positions of other individuals; the best individual is more likely to be learned; the worst individual is more inclined to learn from other individuals.

[0039] Finally, update all individuals by learning the sample vectors, and the update formula is as follows: .

[0040] .

[0041] Among them, represents the position of the current individual under BLS; represents the i th individual's position updated by the sample vector; is a randomly generated [0, 1] vector of dimension dim, representing individuals sorted according to fitness values.

[0042] (3)Sort to obtain the optimal three individuals and update them separately through HS.

[0043] HS strengthens the association between individuals by separately updating the optimal three individuals and allowing other individuals to learn from them. The update formula is as follows: .

[0044] .

[0045] .

[0046] .

[0047] Among them, is the individual ranked first after sorting, named Individual A; is the global optimum at the current iteration; and are two randomly selected individuals and ; represents the current iteration number; represents the maximum number of iterations; is the individual ranked second after sorting, named B the j -th dimensional value of Individual represents a randomly selected dimension; is the individual ranked third after sorting, named C the j -th dimensional value of Individual and respectively represent the j -th dimensional values of the two randomly selected individuals, and ; respectively represent the k -th dimensional values of Individuals A, B, and C; represents a random number with a value range in [0, 1].

[0048] (4)Execute the main part of HLOA.

[0049] The main part of the original HLOA algorithm randomly selects one of the three update methods to update individuals based on a random number, and then replaces the current worst individual. The model is as follows: .

[0050] .

[0051] Among them, represents the i-th individual; are four randomly selected individuals and are not equal to each other; , are random numbers in [0, 1] respectively; , , are respectively set to 1, 1E-6, 0.009807; is a random number generated from a standard Cauchy distribution; represents the current worst individual; and are two random numbers in [0, 0.4046661]; and are two random numbers in [0.5440510, 1]; and are respectively two values randomly selected from the standard color palette, and , the normalized values of each color in the standard color palette are shown in Table 1, is a random number with a value range of [-1, 1].

[0052] Table 1 Normalized values of the standard color palette.

[0053]

[0054] (5) Update the individual through OBL.

[0055] Generalized Opposition-Based Learning (OBL) finds the opposite position of the individual and determines whether to retain the opposite point through fitness value evaluation, thus expanding the exploration range and enhancing the exploration ability. The specific formula is as follows: .

[0056] Among them, represents the opposite point, x represents the current position of the individual.

[0057] (6) Update the individual position through RF.

[0058] The RF strategy learns from the previously updated individual and iteratively learns to enhance the inter-individual correlation. Different from traditional methods, the RF strategy is not limited to the current optimum and can help individuals jump out of local optima. The specific formula is as follows: .

[0059] .

[0060] .

[0061] Among them, Indicates the i th individual after RF update; is the i random dimension value of the th individual; respectively represent the upper and lower limits of the solution space; represents the random dimension value of the optimal individual; represents the random dimension value of the previous updated individual.

[0062] (7) Convert all individuals through the aforementioned transfer function and calculate the fitness value.

[0063] The present invention performs feature selection on 10 low-dimensional public datasets and 10 high-dimensional public datasets respectively. Each algorithm runs independently 10 times and adopts ten-fold cross-validation. The dataset is evenly divided into 10 parts, one part is randomly selected for the test set, and the remaining parts are used for the training set. The number of individuals (N) of all evaluation algorithms is set to 20, and the maximum number of iterations (Max_iter) is set to 50. Finally, the algorithms are evaluated according to the average fitness value, average error rate, and average feature count. And the Wilcoxon signed-rank test is used for the final result. The following tables show the comparison results of fitness values, classification error rates, and the number of selected features respectively. Among them, "+ / = / -" respectively indicate that bROBEHLOA is superior to, similar to, and inferior to this algorithm. See Tables 2, 3, and 4 in the following table.

[0064] bROBEHLOA is respectively compared with the binary horned lizard optimization algorithm (bHLOA), binary gravitational search algorithm (bGSA), binary particle swarm algorithm (bPSO), binary ant lion optimizer (bALO), binary bat algorithm (bBA), binary salp swarm algorithm (bSSA), binary slime mold algorithm (bSMA), binary Harris hawk optimization algorithm (bHHO), and binary moth-flame optimization algorithm (bMFO). Among them, bHLOA is the original version of bROBEHLOA, that is, bROBEHLOA without RF, OBL, BLS, and HS. bGSA, bPSO, bALO, bBA, bSSA, bSMA, bHHO, bMFO are all proposed feature selection methods and have shown strong performance in multiple datasets and are widely used in the field of optimization algorithm feature selection. Therefore, the present invention selects these algorithms as comparison objects to prove the superiority of the present invention in the field of feature selection.

[0065] Table 2 Comparison of the average fitness values of bROBEHLOA and other algorithms on 20 datasets.

[0066]

[0067] Table 3 Comparison of the average classification error rates of bROBEHLOA and other algorithms on 20 datasets.

[0068]

[0069] Table 4 Comparison of the average number of selected features of bROBEHLOA and other algorithms on 20 datasets.

[0070]

[0071] As Figures 2 to 11 shown, it is the convergence curve result graph of bROBEHLOA and each algorithm on 10 low-dimensional public datasets. Figures 12 to 21 Shown is the convergence curve result graph of bROBEHLOA and each algorithm on 10 high-dimensional public datasets.

[0072] The beneficial effects of the present invention are as follows: The present invention proposes an improved horned lizard optimization algorithm based on multiple strategy mechanisms. By using multiple strategy mechanisms, the exploration ability of the horned lizard algorithm is strengthened, the convergence speed is accelerated, and the inter-individual correlation is enhanced. Finally, the effectiveness of bROBEHLOA in feature selection for medical data of different dimensions is demonstrated in medical datasets of different dimensions.

[0073] The embodiment of the present invention provides a feature selection system based on an improved horned lizard optimization algorithm, including: An algorithm improvement module for constructing an improved horned lizard optimization algorithm bROBEHLOA; wherein, the improvement of the algorithm on the horned lizard optimization algorithm includes the following steps: using BLS to obtain the sample vector index of each individual, and updating the individual position according to the sample vector index; sorting the updated individual positions, obtaining the three optimal individuals in the sorting and using the hierarchical structure HS to update the three optimal individuals; using the general opposition-based learning OBL to increase the search range of the algorithm; according to the increased search range, using the random following RF to update the individuals.

[0074] A feature selection module for performing feature selection on the obtained medical dataset, initializing each individual representing the feature subset in the algorithm population; obtaining the classification result of the feature subset using the trained KNN model of the proximity algorithm, and taking the classification accuracy as the fitness value of each individual; using the bROBEHLOA algorithm to find the feature subset with the optimal fitness value from the initialized population, and outputting the optimal feature subset when the maximum number of iterations is reached.

[0075] A specific embodiment is as follows: This embodiment discloses a feature selection method based on an improved horned lizard optimization algorithm, and the specific steps are as follows: S1. Read the medical dataset.

[0076] S2. Limit the data to [-1, 1] and divide the dataset into a test set and a training set through ten-fold cross-validation.

[0077] S3. Initialize the dimension dim, the maximum number of iterations Maxiter, and the optimal vector set Xbest within the parameter range.

[0078] S4. Initialize the algorithm population.

[0079] S5. Train the training set through the KNN model and calculate the fitness value of each individual.

[0080] S6. Update the individual positions using the α-melanocyte-stimulating hormone rate MSH, BLS, hierarchical structure HS, horned lizard optimization algorithm HLOA, general opposition-based learning OBL, and random following RF to construct the improved horned lizard optimization algorithm bROBEHLOA; use bROBEHLOA to find the optimal feature subset.

[0081] S7. Determine whether the current number of iterations exceeds the maximum number of iterations. If not, repeat S5 and S6.

[0082] S8. Output the optimal feature subset.

[0083] The above-described embodiments merely represent several implementation manners of the present invention. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent should be subject to the appended claims.

Claims

1. A feature selection method based on an improved horned lizard optimization algorithm, characterized in that It includes the following steps: Construct an improved horned lizard optimization algorithm bROBEHLOA; wherein, the improvement of the horned lizard optimization algorithm includes the following steps: Use BLS to obtain the sample vector index of each individual, and update the individual position according to the sample vector index; Sort the updated individual positions, obtain the three optimal individuals in the sorting, and use the hierarchical structure HS to update the three optimal individuals; Use the general opposition-based learning OBL to increase the search range of the algorithm; According to the increased search range, use the random following RF to update the individuals; Perform feature selection on the obtained medical dataset, and initialize each individual representing the feature subset in the algorithm population; Use the trained K-nearest neighbor algorithm KNN model to obtain the classification result of the feature subset, and use the classification accuracy as the fitness value of each individual; Use the bROBEHLOA algorithm to find the feature subset with the optimal fitness value from the initialized population, and output the optimal feature subset when the maximum number of iterations is reached.

2. The feature selection method based on the improved horned lizard optimization algorithm according to claim 1, characterized in that The training steps of the K-nearest neighbor algorithm KNN model include: Limit the historical medical dataset to [-1, 1] and divide the dataset into a test set and a training set through ten-fold cross-validation; Use the training set to train the KNN model to obtain the trained KNN model.

3. The feature selection method based on an improved horned lizard optimization algorithm according to claim 1, wherein The steps of using BLS to obtain the sample vector index of each individual and updating the individual position according to the sample vector index are specifically as follows: Obtain the migration rate and emigration rate of each individual: ； Among them, represents the migration rate; represents the removal rate; represents the total number of individuals; Generate a sample vector for each individual according to the migration rate and emigration rate: Update the positions of all individuals according to the sample vector: ； ； in, Indicates the current individual's position under BLS; Indicates i The position of each individual after being updated by the sample vector; is a dim-dimensional randomly generated [0,1] vector; Represents individuals sorted according to their fitness values.

4. The feature selection method based on the improved horned lizard optimization algorithm according to claim 1, wherein The steps of sorting the updated individual positions, obtaining the three optimal individuals in the sorting, and using the hierarchical structure HS to update the three optimal individuals are specifically as follows: Sort all individuals and select the three optimal individuals; Use HS to update the three optimal individuals; ； ； ； ； Among them, The individual ranked first after sorting is named Individual A; Is the global optimum under the current iteration; And Are two randomly selected individuals and ; Represents the current iteration number; Represents the maximum number of iterations; The individual ranked second after sorting is named B The j -th dimension value of the individual; Represents a randomly selected dimension; The individual ranked third after sorting is named C The j -th dimension value of the individual; And Respectively represent the j -th dimension values of the two randomly selected individuals, and ; Respectively represent the k -th dimension values of Individuals A, B, and C; Represents a random number with a value range of [0, 1].

5. The feature selection method based on the improved horned lizard optimization algorithm according to claim 1, characterized in that The steps of using the general opposition-based learning OBL to increase the search range of the algorithm are specifically as follows: Use OBL to obtain the opposite points of the individuals and obtain the fitness values of the opposite points; Determine whether to retain the opposite points according to the fitness values of the opposite points using the greedy idea; ； Among them, represents the opposing point, represents the current position of the individual, represents the upper limit of the solution space, represents the lower limit of the solution space.

6. The feature selection method based on the improved horned lizard optimization algorithm according to claim 1, wherein The steps of using the trained K-nearest neighbor algorithm KNN model to obtain the classification result of the feature subset and using the classification accuracy as the fitness value of each individual are specifically as follows: Convert each individual vector in the training set into a binary vector through the transformation function T function; According to the binary vector, use the trained KNN model to obtain the classification accuracy as the fitness value of each individual.

7. A feature selection system based on an improved horned lizard optimization algorithm, characterized in that It includes: An algorithm improvement module for constructing an improved horned lizard optimization algorithm bROBEHLOA; wherein, the improvement of the horned lizard optimization algorithm includes the following steps: Use BLS to obtain the sample vector index of each individual, and update the individual position according to the sample vector index; Sort the updated individual positions, obtain the three optimal individuals in the sorting, and use the hierarchical structure HS to update the three optimal individuals; Use the general opposition-based learning OBL to increase the search range of the algorithm; According to the increased search range, use the random following RF to update the individuals; A feature selection module is used to perform feature selection on the obtained medical dataset, and initialize each individual representing a feature subset in the algorithm population; use the trained KNN model of the proximity algorithm to obtain the classification result of the feature subset, and take the classification accuracy as the fitness value of each individual; use the bROBEHLOA algorithm to find the feature subset with the optimal fitness value from the initialized population, and output the optimal feature subset when the maximum number of iterations is reached.

Images