High-dimensional abnormal data detection method

Abstract

The invention provides a high-dimensional abnormal data detection method. The objective of the invention is to improve the high-dimensional data anomaly detection effect. The method comprises the following steps: constructing M isolated forests according to position codes of M first ants in the current iteration; based on high-dimensional abnormal data detection, evaluating the effectiveness of the M isolated forests, and determining fitness values corresponding to the M first ants of the current iteration; determining elite ants of the current iteration according to the fitness values corresponding to the M first ants of the current iteration and the fitness values corresponding to the M second ants of the previous iteration; iterating for multiple times according to the steps, and determining a final target position code according to the position code corresponding to the elite ant determined by the last round of iteration; and according to the final target position code, constructing an anomaly detection model, and based on a data segmentation mode corresponding to the anomaly detection model, segmenting the high-dimensional data set to obtain an anomaly detection result.

Description

Technical Field

[0001] The present disclosure relates to the technical field of abnormal data detection, and in particular to a high-dimensional abnormal data detection method. Background Art

[0002] With the rapid development of big data and artificial intelligence technologies and their widespread application in all walks of life, the amount of data has grown rapidly. A large amount of high-dimensional data usually contains redundant or irrelevant features, which greatly increases the difficulty of applying data processing and data mining algorithms to high-dimensional data. On the other hand, abnormal data, or outlier data, is a relatively common phenomenon in different industries. Due to the existence of these data, the overall statistical distribution characteristics of the data are affected and the data quality is reduced, which makes the performance of the model obtained by training the application learning algorithm on the data degraded or even unusable, which may lead to erroneous conclusions.

[0003] Therefore, improving data quality through data cleaning has become an important step before data analysis applications, and how to detect abnormal data is the key to accurate and efficient data cleaning. Among them, with the increasing amount of high-dimensional data, there is a wider demand for effective anomaly detection in high-dimensional data. However, the complex feature space makes anomaly detection significantly more difficult, and how to effectively improve the effect of anomaly detection in high-dimensional data is still a major technical difficulty. Summary of the invention

[0004] In order to overcome the problems existing in the related art, the present disclosure provides a high-dimensional abnormal data detection method. The technical solution of the present disclosure is as follows: According to a first aspect of an embodiment of the present disclosure, a high-dimensional abnormal data detection method is provided, comprising: Determine the position codes of the M first ants in this round of iteration, where the position code of each ant includes: a discrete code group for determining an isolated tree feature subset, and a continuous code group for determining an isolated tree abnormality threshold and an isolated tree weight, where each feature in the isolated tree feature subset represents a segmentation method for a high-dimensional data set; Constructing M isolation forests according to the position codes of the M first ants in the current iteration; Based on high-dimensional abnormal data detection, the effectiveness of the M isolation forests is evaluated respectively, and the fitness value corresponding to each of the M first ants in the current iteration is determined, and the fitness value represents the detection effect of abnormal data in the high-dimensional data set; Determine the elite ant of this iteration according to the fitness values ​​corresponding to the M first ants of this iteration and the fitness values ​​corresponding to the M second ants of the previous iteration, wherein the elite ant is used to determine the position codes of the M first ants of the next iteration; Perform multiple iterations according to the above steps, and determine the final target position code based on the position code corresponding to the elite ant determined in the last round of iteration. The target position code includes: optimal feature subset selection, isolated tree anomaly threshold and isolated tree weight; According to the final target position code, an anomaly detection model is constructed, and based on the data segmentation method corresponding to the anomaly detection model, the high-dimensional data set is segmented to obtain an anomaly detection result.

[0005] Optionally, determining the elite ant of the current iteration according to the fitness values ​​corresponding to the M first ants of the current iteration and the fitness values ​​corresponding to the M second ants of the previous iteration includes: Sorting the M first ants in the current iteration and the M second ants in the previous iteration according to the fitness values ​​corresponding to the M first ants in the current iteration and the fitness values ​​corresponding to the M second ants in the previous iteration to obtain a first sorting result; Determine the first M ants in the first sorting result as the M second ants in this round of iteration; The ant with the highest fitness among the M second ants in the current iteration is determined as the elite ant in the current iteration.

[0006] Optionally, it also includes: Sorting the M first ants in the current iteration according to the fitness values ​​corresponding to the M first ants in the current iteration to obtain a second sorting result; Taking half of the ants with high fitness values ​​in the second sorting result corresponding to the M first ants in the current iteration as a better ant pool, and taking half of the ants with low fitness values ​​in the second sorting result as a worse ant pool; According to a preset matching number, ants are selected from the better ant pool and the worse ant pool respectively to obtain a matching number of matching ant groups; each matching ant group includes an ant from the better ant pool and an ant from the worse ant pool; For each matching ant group, the two ants included in the group exchange their continuous codes to obtain new ants; According to the fitness values ​​corresponding to the M first ants of the current iteration and the fitness values ​​corresponding to the M second ants of the previous iteration, the M first ants of the current iteration and the M second ants of the previous iteration are sorted to obtain a first sorting result, including: The new ants, the M first ants in the current iteration, and the M second ants in the previous iteration are sorted according to their respective corresponding fitness values ​​to obtain a first sorting result.

[0007] Optionally, constructing M isolation forests according to the position codes of the M first ants in the current iteration includes: constructing an isolation forest according to the position code of each first ant; Each isolation forest is constructed as follows: Decoding each discrete code in the discrete code group that is sorted in sequence to obtain each feature that is sorted in sequence; According to the order of each feature, determine the feature subset of each isolated tree in the isolated forest corresponding to the first ant; Decoding the first continuous code in the continuous code group to obtain the isolated tree abnormality threshold of the first ant; Decoding the remaining sequentially ordered continuous codes in the sequentially ordered group to obtain respective weight values ​​that are sequentially ordered; Determine the weight of each isolated tree in the isolated forest corresponding to the first ant according to the order of each weight value; An isolated forest corresponding to the first ant is constructed according to the feature subsets corresponding to the isolated trees, the weights corresponding to the isolated trees, and the isolated tree abnormality threshold.

[0008] Optionally, determining the position codes of the M first ants in this round of iteration includes: According to the elite ants obtained in the previous round of iteration, the target ant selects feature subsets corresponding to each isolated tree in the isolated forest from the feature selection space, wherein the feature subsets include multiple features; wherein the feature subsets are sorted in order, and each feature in the feature subsets is sorted in order, and the target ant is any one of the M second ants in the previous round of iteration; Encoding the features in the feature subset according to a first encoding format to obtain discrete codes of the features in the discrete code group; The target ant updates the position vector of the target ant according to the wandering range and wandering step length of this iteration and based on the elite ant obtained in the previous iteration; According to the position vector of the target ant, encoding is performed according to a second encoding format to obtain a continuous encoding group corresponding to the target ant; The position code of any one of the M first ants in the current iteration is determined according to the discrete code group composed of the discrete codes and the continuous code group.

[0009] Optionally, based on the elite ants obtained in the previous iteration, the target ants select feature subsets corresponding to each isolated tree in the isolation forest from the feature selection space, including: According to the discrete coding group corresponding to the elite ants obtained in the previous iteration, the pheromone concentration of each feature in the feature selection space in this iteration is determined; Determine the number of selection steps of the target ant, where the number of selection steps is used to determine the number of features included in the feature subset; In the case where the target ant selects a feature from the optional features in the feature search space at each step, determining the selection probability of each optional feature according to the pheromone concentration and heuristic information of each optional feature; the optional feature represents the feature that is not selected; Based on the selection probability of the optional feature, select an optional feature as the feature corresponding to the step according to the roulette strategy; According to the features corresponding to each step, determine the feature subset of the isolated tree; After determining a feature subset of an isolated tree, the selected state of the features in the feature selection space is cleared.

[0010] Optionally, when the target ant selects a feature from the optional features in the feature search space at each step, determining the selection probability of each optional feature according to the pheromone value and heuristic information of each optional feature includes: Determine a multiple selection coefficient of the optional feature, wherein the multiple selection coefficient is used to control the diversity in the feature selection process; a value corresponding to the multiple selection coefficient of the feature is determined according to the number of times the feature is selected in this round of iteration; The selection probability of each optional feature is determined according to the multiple selection coefficients, pheromone values ​​and heuristic information of the optional features.

[0011] Optionally, according to the discrete coding group corresponding to the elite ants obtained in the previous iteration, the pheromone concentration of each feature in the feature selection space in the current iteration is determined, including: Determine a feature subset of the elite ant according to the discrete coding group corresponding to the elite ant obtained in the previous iteration; Performing pheromone enhancement processing on the features of the feature subset corresponding to the elite ants, and performing pheromone volatilization processing on the features of the feature subsets corresponding to the remaining ants obtained in the previous round of iterations; The pheromone concentration of each feature in the feature selection space of this round of iteration is determined through the pheromone enhancement process and the pheromone volatilization process.

[0012] Optionally, the target ant updates the position vector of the target ant according to the wandering range and wandering step length of this iteration based on the elite ant obtained in the previous iteration, including: Selecting a roulette ant from the M second ants of the previous iteration in a roulette manner; According to the wandering range and wandering step length of this round of iteration, the target ant respectively surrounds the elite ant and the roulette ant obtained in the previous round of iteration to determine the elite position vector and the roulette position vector; The position vector of the target ant is updated according to the elite position vector and the roulette position vector.

[0013] Optionally, the walking step length and the walking range are determined by the following steps: Determine preset upper and lower bounds of the dimension of the target ant; Determine the wandering range of the target ant according to the upper and lower bounds of the dimension of the target ant, the current iteration number corresponding to this round of iteration, and the maximum iteration number; According to the current iteration number corresponding to the current iteration, the walking step length of the target ant in the current iteration is determined.

[0014] According to a second aspect of an embodiment of the present disclosure, a high-dimensional abnormal data detection device is provided, comprising: A determination module is used to determine the position codes of the M first ants in this round of iteration, and the position code of each ant includes: a discrete code group for determining an isolated tree feature subset, and a continuous code group for determining an isolated tree abnormality threshold and an isolated tree weight, wherein each feature in the isolated tree feature subset represents a segmentation method for a high-dimensional data set; A construction module, used to construct M isolated forests according to the position codes of the M first ants in the current iteration; An evaluation module is used to evaluate the effectiveness of the M isolation forests based on high-dimensional abnormal data detection, and determine the fitness value corresponding to each of the M first ants in the current iteration, wherein the fitness value represents the detection effect of abnormal data in the high-dimensional data set; An updating module, used to determine the elite ants of this round of iteration according to the fitness values ​​corresponding to the M first ants of this round of iteration and the fitness values ​​corresponding to the M second ants of the previous round of iteration, wherein the elite ants are used to determine the position codes of the M first ants of the next round of iteration; An iteration module is used to perform multiple iterations according to the above steps, and determine the final target position code according to the position code corresponding to the elite ant determined in the last round of iteration. The target position code includes: optimal feature subset selection, isolated tree abnormality threshold and isolated tree weight; The end module is used to build an anomaly detection model according to the final target position code, and to segment the high-dimensional data set based on the data segmentation method corresponding to the anomaly detection model to obtain an anomaly detection result.

[0015] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the high-dimensional abnormal data detection method described in the first aspect are implemented.

[0016] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the high-dimensional abnormal data detection method described in the first aspect are implemented.

[0017] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps of the high-dimensional abnormal data detection method as described in the first aspect are implemented.

[0018] The present invention optimizes the parameters of isolated trees in the isolation forest algorithm through the position coding of ants, which can effectively improve the anomaly detection effect of high-dimensional data using isolation forests, and can overcome the problem that the traditional isolation forest has insufficient anomaly detection ability for high-dimensional anomaly data. The position coding of each ant includes two parts, discrete coding and continuous coding. This coding method enables ants to search and optimize in two different spaces (discrete space and continuous space). Through this dual coding method, the present invention can simultaneously perform feature selection and model parameter optimization, thereby improving the effect of anomaly detection. The M isolated forests determined in each round of iteration are evaluated for the abnormal data detection effect on the high-dimensional data set to obtain the evaluation results; according to the evaluation results, the ants with the highest fitness values ​​among the M first ants in this round of iteration and the M second ants in the previous round of iteration are selected as elite ants, which can retain the information of the M second ants in the previous round of iteration, ensure that the elite ants determined in this round of iteration are the best ants, and guide the next round of iteration according to the elite ants, which can help the ant group to get closer to a better solution space. Through multiple iterations, the ant group continuously adjusts the position encoding according to the guidance of the elite ants updated in the previous round, gradually optimizes the feature subset, threshold and weight, so that the model can continuously approach the global optimal solution (that is, continuously determine the optimal high-dimensional data set segmentation method), and obtain anomaly detection model, which contains the optimal high-dimensional data set segmentation method. Therefore, by detecting abnormal data in the high-dimensional data set through the anomaly detection model, the abnormal data in the high-dimensional data set can be efficiently identified, thereby improving the accuracy and stability of anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the description of the embodiments of the present disclosure will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 is a schematic diagram of the steps of a high-dimensional abnormal data detection method shown in an embodiment of the present disclosure; Figure 2 is a schematic diagram of position coding shown in an embodiment of the present disclosure; Figure 3 is a schematic diagram of a feature selection process shown in an embodiment of the present disclosure; Figure 4 It is a schematic diagram of the steps of a high-dimensional abnormal data detection algorithm shown in an embodiment of the present disclosure; Figure 5 is a schematic diagram of a high-dimensional abnormal data detection device shown in an embodiment of the present disclosure Figure 6 It is a schematic diagram of an electronic device shown in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0022] The terms "first", "second", etc. in the specification and claims of the present disclosure are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable when appropriate, so that the embodiments of the present disclosure can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0023] First, the professional terms involved in this disclosure are explained: Isolation forest: Isolation forest does not need to use distance, density and other indicators to measure the differences between samples. Instead, it constructs an isolation tree to perform binary tree splitting on the data set samples to characterize the degree of alienation (isolation) between the abnormal samples and the main trunk of the tree structure. It is easy to understand, simple and efficient. The final decision result of the isolation forest is obtained by integrating the decision results of several isolation trees. The threshold of the isolation tree, that is, the proportion or number of abnormal data in the training data, and the integration weight parameter of the isolation tree are important parameters for training and integrating the isolation tree, which have a key impact on the performance of the final isolation forest model.

[0024] Ant colony optimization (ACO): Ant colony optimization is an evolutionary algorithm inspired by the behavior of ants in nature in finding food. It simulates the natural behavior of ants releasing pheromones in the process of searching for food. In this way, ants can communicate indirectly with each other and cooperate to find the shortest path from the ant nest to the food source.

[0025] Ant lion optimizer (ALO): The ant lion optimizer simulates the behavior of ant lions in nature when preying on ants: the ant lion digs a cone-shaped pit in the sand and hides at the bottom of the cone waiting for prey; when the ant enters the trap, the ant lion will throw sand to the edge to make the ant slide down the cone to the bottom of the pit; when the ant slides to the bottom of the pit, the ant lion will prey on the ant; then it will repair the trap and prepare for the next hunt.

[0026] With the rise of big data, the scale of data has increased significantly. The amount of data generated daily has rapidly jumped from PB (Petabyte) and EB (Exabyte) levels to ZB (Zettabyte) and even YB (Yottabyte). High dimensionality is an important feature of a large amount of data. This high-dimensional data usually contains redundant or irrelevant features, which increases the complexity of data processing and data analysis.

[0027] Isolation forests can be used to detect abnormal data. However, the classic isolation forest method uses all features to train isolated trees, which has low applicability to high-dimensional data. In order to solve this problem, the present disclosure provides a high-dimensional abnormal data detection method, which can reduce the feature space complexity of isolated trees and improve the diversity of isolated trees.

[0028] Figure 1 Schematic diagram of a method for detecting high-dimensional abnormal data according to an embodiment of the present disclosure. Figure 1As shown, the method may specifically include the following steps S11 to S16.

[0029] Step S11: Determine the position codes of the M first ants in this round of iteration, and the position code of each ant includes: a discrete code group for determining an isolated tree feature subset, and a continuous code group for determining an isolated tree anomaly threshold and an isolated tree weight, wherein each feature in the isolated tree feature subset represents a way of segmenting a high-dimensional data set.

[0030] The input of each iteration is the M second ants of the previous iteration. The M second ants of the previous iteration are the M ants with the best detection effect for high-dimensional abnormal data in the previous iteration. The M second ants can update their corresponding position codes in this iteration. The M second ants of the previous iteration that update the position codes are the M first ants of this iteration. Ants with different position codes are different ants.

[0031] The position coding of ants consists of two parts, including a discrete coding group and a continuous coding group. M is an integer greater than 0.

[0032] The discrete coding group is used to determine the feature subset corresponding to each isolated tree in the isolation forest. Each isolated tree corresponds to a feature subset. When there are I isolated trees, the discrete coding group includes I feature subsets. When the isolation forest detects abnormal data in a high-dimensional data set, each isolated tree selects a feature from its corresponding feature subset as a segmentation method to segment the data in the high-dimensional data set, thereby obtaining the abnormal data that the isolated tree can determine. Each feature included in the feature subset corresponding to the isolated tree represents a segmentation method of one dimension. For example, when the feature subset corresponding to an isolated tree includes feature A, feature B, and feature C, feature A, feature B, and feature C each represent a segmentation method. When the isolated tree selects feature A, the high-dimensional data set is segmented based on the segmentation method corresponding to feature A.

[0033] By establishing feature subsets for isolated trees, the feature space complexity of isolated trees can be effectively reduced, avoiding the isolated trees from selecting features as a splitting method from all the features corresponding to the high-dimensional data set, but instead selecting features as a splitting method from a limited feature subset; in addition, constructing different feature subsets for different isolated trees can also effectively improve the diversity of isolated trees.

[0034] Discrete coding means which features are selected from all features as the features of any isolated tree, so as to construct the feature subset of the isolated tree. The continuous coding group is used to determine the abnormal threshold of the isolated tree and the weight of the isolated tree. The abnormal threshold and the weight of the isolated tree can be used to distinguish abnormal data from normal data; among them, the abnormal threshold of the isolated tree can be used to determine the proportion or number of abnormal data in the data set, and the weight of the isolated tree is used to obtain the final decision result of the isolated forest on abnormal data.

[0035] Step S12: construct M isolated forests according to the position codes of the M first ants in the current iteration.

[0036] In each iteration, M isolation forests are constructed based on the position codes of the M first ants obtained in this iteration. Each ant can generate an isolation forest that matches its code, that is, each ant represents a candidate anomaly detection model. The feature subset of the isolated tree in the isolation forest is determined by the discrete coding part, and the anomaly threshold and weight are determined by the continuous coding part.

[0037] Step S13: Based on high-dimensional abnormal data detection, the effectiveness of the M isolation forests is evaluated respectively, and the fitness value corresponding to each of the M first ants in the current iteration is determined, and the fitness value represents the detection effect of abnormal data in the high-dimensional data set.

[0038] Based on high-dimensional abnormal data, the performance of each isolation forest is evaluated, specifically, the anomaly detection effect of the isolation forest on the high-dimensional data set is evaluated. The anomaly data in the high-dimensional data set is detected using the isolation forest to determine the anomaly data detection effect of the isolation forest, thereby evaluating the effectiveness of the M isolation forests. An isolation forest with high effectiveness means that it can effectively identify anomaly data in the high-dimensional data set.

[0039] By evaluating the effectiveness of the isolation forest corresponding to each first ant, a fitness value can be assigned to each first ant, which indicates the pros and cons of the model represented by the ant in the current high-dimensional anomaly data detection. The fitness value reflects the performance of the corresponding isolation forest. The higher the fitness value of the ant, the better the effect of the isolation forest corresponding to the position encoding of the ant in high-dimensional anomaly data detection.

[0040] Step S14: Determine the elite ant of this iteration according to the fitness values ​​corresponding to the M first ants of this iteration and the fitness values ​​corresponding to the M second ants of the previous iteration, and the elite ant is used to determine the position code of the M first ants of the next iteration.

[0041] The M second ants in the previous round of iteration are the input of this round of iteration. The M second ants are selected in the previous round of iteration based on the fitness corresponding to each ant. Therefore, when the M second ants in the previous round of iteration are input into this round of iteration, the M second ants carry their corresponding fitness values ​​and do not need to be calculated again in this round of iteration.

[0042] In each round of iteration, after determining the fitness values ​​corresponding to the M first ants of this round of iteration, the elite ants will be updated according to the fitness values ​​corresponding to the M first ants of this round of iteration and the M second ants of the previous round of iteration. The elite ants represent the best model in the current iteration, that is, the best feature selection and parameter configuration, which includes the isolated tree anomaly threshold and isolated tree weight. The position encoding of the M first ants represents the solution space explored in the current iteration. By introducing the M second ants of the previous round of iteration, the fitness values ​​of the ants in this round and the previous round can be compared, and individuals with better fitness can be selected as elite ants, ensuring that even if the solution of individual ants in a certain round of iteration performs poorly, it can be supplemented by selecting excellent solutions from the previous round of iteration, thereby avoiding the loss of useful ant information.

[0043] The performance of elite ants provides the best guidance for the next iteration. In each iteration, the position codes of the M first ants in the iteration are determined using the elite ants updated in the previous iteration. For the first iteration, the position codes of the M ants are initialized, and the fitness values ​​of all initialized ants are calculated in advance. The ants with the best fitness values ​​are used as elite ants to guide the determination of position codes in the first iteration.

[0044] Specifically, the position codes of the elite ants can be used as a reference to guide the determination of the position codes of the M first ants in the next iteration. In each round, the position codes of the M ants input in this round (i.e., the M second ants in the previous iteration) are updated based on the elite ants re-determined in the previous iteration, so that the ant colony can explore a better solution space along the direction guided by the elite ants.

[0045] By updating elite ants, we can ensure that the ant colony approaches the optimal solution. Elite ants guide the search direction of the ant colony and effectively avoid falling into local optimality.

[0046] Step S15: perform multiple iterations according to the above steps, and determine the final target position code according to the position code corresponding to the elite ant determined in the last round of iteration. The target position code includes: optimal feature subset selection, isolated tree abnormality threshold and isolated tree weight.

[0047] The above process is repeated for multiple rounds of iterations. In each round of iteration, the position codes of the M ants input in this round of iteration (i.e., the M second ants in the previous round of iteration before this round of iteration) are continuously updated and evaluated.

[0048] The position codes of the M ants are continuously updated in multiple iterations, and new M ants with different position codes are continuously generated. The elite ants determined in each round of iteration will continuously provide guidance to the M ants input in the next round of iteration, so that the M ants updated in each round of iteration can explore the optimal solution space in the next round of iteration and obtain M ants with updated position codes.

[0049] As the iteration proceeds, the position codes of the M first ants and the M second ants will be highly similar, and the algorithm corresponding to the method converges. Based on the position codes of the M first ants determined in the last round of iteration and the position codes of the M second ants input in the last round of iteration (i.e., the M second ants in the previous round of iteration before the last round of iteration), the final target position code is obtained. Specifically, the position code of the elite ant updated in the last round of iteration can be determined as the target position code. The elite ant represents the ant with the best effect in detecting abnormal data in a high-dimensional data set among the M first ants in the last round of iteration and the M second ants input in the last round of iteration. The position code corresponding to the elite ant includes the optimal feature subset selection, isolated tree anomaly threshold and isolated tree weight. Among them, the optimal feature subset selection, isolated tree anomaly threshold and isolated tree weight refer to that the detection effect of abnormal data in a high-dimensional data set meets the preset requirements.

[0050] Step S16: construct an anomaly detection model according to the final target position code, and segment the high-dimensional data set based on the data segmentation method corresponding to the anomaly detection model to obtain an anomaly detection result.

[0051] Based on the target position encoding, an anomaly detection model is constructed. The anomaly detection model detects abnormal data in high-dimensional data sets based on the isolation forest corresponding to the target position encoding.

[0052] In the last round of iteration, the final target position encoding is determined based on the position encodings of the M first ants in the last round of iteration and the M second ants input in the last round of iteration. The target position encoding represents the optimal feature subset selection, anomaly threshold and isolated tree integration weight configuration. This enables: each isolated tree in the isolation forest to more effectively detect abnormal data in high-dimensional data sets based on feature subsets, and the isolation forest integrates the abnormal data detection results of each isolated tree based on the optimal isolated tree anomaly threshold and isolated tree weight to obtain the final abnormal data detection result.

[0053] Based on this optimal configuration, the final anomaly detection model is constructed. The final anomaly detection model is used to perform anomaly detection on the data set.

[0054] By adopting the embodiment of the present disclosure, the dual encoding of the discrete encoding group and the continuous encoding group is combined to realize the parallel processing of feature selection and isolation forest parameter optimization, and the correct feature subset can be selected and the abnormal threshold and weight can be optimized. Each ant can obtain an independent isolation forest, and each isolation forest can be constructed based on different feature subsets and parameter configurations, so that each isolation forest has a different perspective to identify abnormal data. The combination of multiple isolation forests obtained in one round of iteration enhances the robustness and detection ability of the model, so that the present disclosure can detect anomalies on the data set from multiple perspectives. After evaluating the effectiveness of each isolation forest, the elite ants are updated. The feature subset selection and isolation forest parameter configuration of the elite ants provide good guidance for the next round of iteration. Through the guidance of the elite ants, the present disclosure can focus the search process on the area with the greatest hope of finding the global optimal solution, avoiding the problem of falling into the local optimal solution; on the other hand, through the guidance of the elite ants, the M ants will adjust their position codes based on the features and parameters of the elite ants in the next round of iteration, and ensure the efficiency of the search on the basis of the ant colony gradually moving closer to the optimal solution.

[0055] Among them, in an optional embodiment, the elite ants of this round of iteration are determined according to the fitness values ​​corresponding to each of the M first ants of the current round of iteration and the fitness values ​​corresponding to each of the M second ants of the previous round of iteration, including: sorting the M first ants of the current round of iteration and the M second ants of the previous round of iteration according to the fitness values ​​corresponding to each of the M first ants of the current round of iteration and the fitness values ​​corresponding to each of the M second ants of the previous round of iteration to obtain a first sorting result; determining the first M ants in the first sorting result as the M second ants of this round of iteration; and determining the ant with the highest fitness among the M second ants of the current round of iteration as the elite ant of this round of iteration.

[0056] The fitness value represents the quality of the ant's solution in the current task. Specifically, the fitness value reflects the effect of the isolated tree feature subset, threshold and weight selected by the ant in anomaly detection of high-dimensional data sets. By comparing the fitness of the ants in this round and the previous round, it can ensure that the best solutions of the current and previous rounds are effectively integrated, providing more comprehensive information for the subsequent elite ant selection, avoiding the limitation of making decisions based only on the performance of the current round.

[0057] Combine the fitness values ​​of the M first ants in this round with the fitness values ​​of the M second ants in the previous round, and sort them in descending order according to their respective fitness values ​​to obtain the first sorting result. Select the top M ants from the first sorting result and use them as the M second ants in this round of iteration and as the input for the next round of iteration. In the next round of iteration, the position codes of the input M second ants are updated to obtain the M first ants for the next round of iteration. This screening method ensures that the optimal solution (i.e., ants with higher fitness values) in each round of iteration can be retained and passed to the next round. By selecting ants with higher fitness, effective exploration of the search space is guaranteed and possible "information loss" is reduced.

[0058] The elite ant represents the optimal solution of the current iteration. The corresponding isolated tree feature subset, anomaly threshold and weight of the solution can optimize the abnormal data detection effect to the greatest extent. By selecting the ant with the best fitness as the elite, it is ensured that the model converges to the optimal solution after each iteration. Therefore, the ant with the highest fitness value among the M second ants in this iteration is determined as the elite ant of this iteration.

[0059] By adopting the embodiments of the present disclosure, by combining the M first ants of this round of iteration and the M second ants of the previous round of iteration, the M second ants of this round of iteration with the best fitness are sorted and selected, thereby ensuring the effective retention and optimization of the optimal solution in each round of iteration, so that the ant algorithm can continuously converge to the global optimal solution, which not only improves the search efficiency, but also avoids information loss, and ensures the accuracy and stability of high-dimensional abnormal data detection. The ant colony is updated by eliminating the strategy of low fitness values. This dynamic update mechanism ensures that the ant colony can continue to evolve and approach a better solution. Selecting the ant with the highest fitness value from the updated ant colony as the elite ant helps to avoid the algorithm from falling into a local optimal solution and ensures that the algorithm can continue to approach the global optimal solution. Among them, in an optional embodiment, it also includes: sorting the M first ants in the current iteration according to the fitness values ​​corresponding to each of the M first ants in the current iteration to obtain a second sorting result; using half of the ants with high fitness values ​​in the second sorting result corresponding to the M first ants in the current iteration as a better ant pool, and using half of the ants with low fitness values ​​in the second sorting result as a worse ant pool; selecting ants from the better ant pool and the worse ant pool respectively according to a preset matching number to obtain a matching number of matching ant groups; each matching ant group includes an ant in the better ant pool and an ant in the worse ant pool. ants; exchanging continuous codes of two ants included in each matching ant group to obtain new ants; sorting the M first ants of the current iteration and the M second ants of the previous iteration according to the fitness values ​​corresponding to each of the M first ants of the current iteration and the fitness values ​​corresponding to each of the M second ants of the previous iteration to obtain a first sorting result, including: sorting the new ants, the M first ants of the current iteration and the M second ants of the previous iteration according to the fitness values ​​corresponding to each of the new ants, the M first ants of the current iteration and the M second ants of the previous iteration to obtain a first sorting result.

[0060] In this round of iteration, after obtaining the fitness values ​​corresponding to the M first ants of this round of iteration, the M first ants of this round of iteration can be sorted in order from high to low according to the fitness values ​​to obtain a second sorting result, and half of the ants with high fitness values ​​in the second sorting result are classified as a better ant pool; and half of the ants with low fitness values ​​are classified as a worse ant pool.

[0061] According to the preset matching number, ants are selected from the better ant pool and the worse ant pool to obtain the corresponding number of matching ant groups. For example, when the matching number is 2, 2 matching ant groups are required. For each matching ant group, one ant is selected from the better ant pool and the worse ant pool to match and obtain the matching ant group. By pairing better and worse ants, it is ensured that the worse ants can learn from the advantages of the better ants, thereby achieving the optimization of the worse ants.

[0062] The continuous codes of the two ants in each matching ant group are exchanged, that is, the parts of the two ants used to determine the abnormal threshold of the isolated tree and the weight of the isolated tree are exchanged. After the exchange, new ants are obtained. The number of new ants obtained in each matching ant group is consistent with the number of ants included in the matching ant group, that is, two new ants. For example, a matching ant group includes an inferior ant A and a superior ant B. The position code of the inferior ant A includes discrete code A and continuous code A, and the position code of the superior ant B includes discrete code B and continuous code B. After exchanging the continuous code parts of the inferior ant A and the superior ant B, new ants C and new ants D are obtained. The position code of the new ant C includes discrete code A and continuous code B, and the position code of the new ant D includes discrete code B and continuous code A. By allowing the superior and inferior ants to exchange the continuous code parts, the excellent features of the superior ant can be transferred to the inferior ant, thereby improving the performance of the inferior ant.

[0063] After the new ants are generated, based on the fitness values ​​of the new ants, the M first ants of this round of iteration, and the M second ants of the previous round of iteration, the new ants, the M first ants of this round of iteration, and the M second ants of the previous round of iteration are sorted in order from high to low according to the fitness values ​​to obtain a first sorting result.

[0064] By using the embodiment of the present disclosure, the M first ants of this round of iteration are sorted according to the fitness value, and the ants can be clearly divided into two groups with better performance and worse performance. The better ants exchange information with the worse ants (i.e., exchange continuous coding parts), which helps the worse ants learn the excellent features of the better ants, thereby improving the quality of the overall ant colony and accelerating the convergence process of the algorithm. In each round of iteration, the first sorting result is determined according to the fitness values ​​of the newly generated ants, the M first ants of this round of iteration, and the M second ants of the previous round of iteration.

[0065] Among them, in an optional embodiment, M isolated forests are constructed according to the position codes of the M first ants in the current iteration, including: constructing an isolated forest according to the position code of each first ant; each isolated forest is constructed in the following manner: decoding each discrete code sorted in sequence in the discrete code group to obtain each feature sorted in sequence; determining the feature subset of each isolated tree in the isolated forest corresponding to the first ant according to the order of each feature; decoding the first continuous code in the continuous code group to obtain the isolated tree anomaly threshold of the first ant; decoding the remaining continuous codes sorted in sequence in the continuous code group to obtain each weight value sorted in sequence; determining the weight of each isolated tree in the isolated forest corresponding to the first ant according to the order of each weight value; constructing the isolated forest corresponding to the first ant according to the feature subset corresponding to each isolated tree, the weight corresponding to each isolated tree and the isolated tree anomaly threshold.

[0066] The structure of the position coding can be determined according to the anomaly detection model to be constructed. Specifically, when the isolation forest corresponding to the anomaly detection model needs to include I isolated trees and the feature subset corresponding to each isolated tree needs to include q features, the position coding needs to include discrete codes corresponding to I*q features, continuous codes corresponding to I isolated tree weights, and continuous codes corresponding to an isolated tree anomaly threshold.

[0067] Figure 2 is a schematic diagram of position coding shown in an embodiment of the present disclosure. Figure 2 As shown, each discrete code represents a feature selected to enter the feature subset. When the isolation forest includes I isolated trees and the feature subset corresponding to each isolated tree needs to include q features, the discrete code 1-discrete code q in the discrete code group represents the q features included in the feature subset corresponding to one of the isolated trees in the isolation forest. The continuous code group includes two parts, including an isolated tree threshold code representing the abnormal threshold of the isolated tree and an isolated tree weight code representing the weight of the isolated tree. In the continuous code part, the abnormal threshold of the isolated tree is represented by a continuous code, and each isolated tree can distinguish normal data from abnormal data through the continuous code. The number of isolated tree weight codes in the continuous code part is consistent with the number of isolated trees in the isolation forest. When there are I isolated trees in the isolation forest, I isolated tree weight codes are required in the discrete code group.

[0068] The position of each ant can be encoded as Figure 2 Each ant can construct a corresponding isolated forest. For the position encoding of the M first ants in this round of iteration, M corresponding isolated forests can be constructed.

[0069] For each discrete code group of the first ant, first decode it in the order of coding to obtain a series of features. Each discrete code represents a specific feature. By decoding the discrete codes in the discrete code group, the features used by each isolated tree can be determined. The number of features decoded from the discrete code group needs to meet the number of features required by the isolated forest. For example, when the isolated forest includes I isolated trees and the feature subset corresponding to each isolated tree needs to include q features, the number of features decoded from the discrete code group needs to meet I*q.

[0070] Decoding the continuous code group of each first ant can obtain an abnormal threshold and several weight values. Specifically, decoding the first continuous code in the continuous code group can obtain the isolated tree abnormal threshold corresponding to the isolated forest, and continue to decode the remaining continuous codes to obtain a series of weight values.

[0071] For a series of features and a series of weight values ​​obtained by position encoding, the feature subset and weight used by each isolated tree are determined. The feature order determines which features will be applied to the isolated tree, and the weight value determines the influence and importance of each isolated tree in the isolation forest.

[0072] Specifically, the features corresponding to the same isolated tree are connected head to tail in the features obtained by decoding and arranged in sequence. The series of features obtained by decoding and arranged in sequence are sequentially assigned to different isolated trees from the first feature to the last feature, wherein, after the feature subset of an isolated tree is determined, the remaining features are sequentially assigned to the next isolated tree, thereby obtaining the feature subsets of the remaining next isolated trees.

[0073] The series of weight values ​​obtained by decoding are assigned to different isolated trees in sequence.

[0074] After decoding a position code and assigning the decoded features, isolated tree anomaly thresholds, and isolated tree weights to each isolated tree in the isolation forest, an isolation forest corresponding to the position code can be constructed based on the assignment results.

[0075] By using the embodiments of the present disclosure, the isolated forest constructed by each ant will have different feature selections, weights, and anomaly thresholds, so that the models of different ants differ in their ability to detect abnormal data, ensuring that the present disclosure can be optimized simultaneously in multiple dimensions, enhancing the exploratory nature of the swarm algorithm, and allowing different ants to find the optimal solution in different local solution spaces. The decoding process of the discrete coding group determines the features used by the isolated trees in the isolated forest, so that each isolated tree can customize a feature subset for a specific data set and problem, avoiding interference from redundant features and improving the efficiency of the model. By setting different anomaly thresholds and weights for each isolated tree, the contribution of each isolated tree can be adjusted, so that the model can better adapt to changes in high-dimensional complex data and improve detection results.

[0076] In one optional embodiment, the position codes of the M first ants in this iteration are determined, including: according to the elite ants obtained in the previous iteration, the target ant selects a feature subset corresponding to each isolated tree in the isolated forest from the feature selection space, wherein the feature subset includes multiple features; wherein the feature subset is sorted in order, and each feature in the feature subset is sorted in order, and the target ant is any one of the M second ants in the previous iteration; the feature in the feature subset is encoded according to a first encoding format to obtain a discrete code of the feature in the discrete coding group; the target ant updates the position vector of the target ant according to the wandering range and wandering step length of this iteration and based on the elite ants obtained in the previous iteration; according to the position vector of the target ant, encoding is performed according to the second encoding format to obtain a continuous coding group corresponding to the target ant; according to the discrete coding group composed of the discrete codes and the continuous coding group, the position code of any one of the M first ants in this iteration is determined.

[0077] The elite ants obtained in the previous iteration will affect the pheromones of each feature in the feature selection space. Specifically, when a feature is selected into the discrete coding group of the elite ants, the feature will carry a more concentrated pheromone in the feature selection space. When the target ants select features from the feature selection space, they select based on the pheromones carried by the features. The more concentrated the pheromone carried by a feature, the greater the probability that the feature will be selected.

[0078] The elite ants determined in the previous iteration can affect the feature selection of the M ants input to this iteration (i.e., the M second ants in the previous iteration). The features selected based on the elite ants determined in the previous iteration can be closer to the optimal solution. The target ant is any one of the M second ants in the previous iteration. Taking the target ant as an example, this embodiment describes how the M second ants in the previous iteration update their position codes in this iteration to obtain the M first ants in this iteration.

[0079] When the target ant determines the feature subsets for each isolated tree in the corresponding isolated forest, the feature subsets are determined in the order of the isolated trees. After the feature subset of an isolated tree is determined, the feature subset of the next isolated tree is determined, and the features selected by each ant are also sorted in the order of selection. Among them, the feature subset corresponding to each isolated tree is determined from the complete feature selection space.

[0080] The selected features are encoded according to the first encoding format to obtain Figure 2 The discrete codes shown in , each selected feature corresponds to a discrete code. All discrete codes are combined in the order in which their corresponding features are selected, and the discrete code group of the target ant is updated.

[0081] In the part of determining the isolated tree abnormality threshold and the isolated tree weight, the present disclosure adopts the ALO search method.

[0082] The position of the target ant in this iteration will be updated based on the performance of the elite ant in the previous iteration. The target ant has a position vector that represents the position in the continuous search space. The target ant will update the position vector corresponding to its current position based on the elite ant according to its wandering range and step length. Specifically, the target ant wanders around the elite ant in the wandering range according to the wandering step length, thereby updating its position. The target ant updates the target ant's continuous coding group based on the updated position vector.

[0083] As the iteration progresses, the range and step length of each iteration will change. The range and step length determine the distance and direction of the ant's movement in the search space. Specifically, as the iteration progresses, the ant's range will shrink, and in each iteration, the ant will recalculate a walking step length.

[0084] By updating the continuous coding group and the discrete coding group of the target ant, the position coding of the target ant is updated, and the target ant after the position coding is updated will be used as one of the M first ants in this round of iteration.

[0085] Among them, in an optional embodiment, the walking step length and the walking range are determined by the following steps: determining the preset upper and lower bounds of the dimension of the target ant; determining the walking range of the target ant according to the upper and lower bounds of the dimension of the target ant, the current number of iterations corresponding to this round of iterations, and the maximum number of iterations; determining the walking step length of the target ant in this round of iterations according to the current number of iterations corresponding to this round of iterations.

[0086] As the iterations progress, the ants’ roaming range will gradually shrink, prompting the algorithm to gradually focus on the refined search for the solution. The maximum number of iterations determines the overall running time and search depth of the algorithm. The roaming range should be adjusted according to the different stages of the maximum number of iterations so that the ants’ search behavior is adapted to the iteration progress.

[0087] The target ant's wandering range in each iteration can be determined by the following formula:

[0088] c and d represent the upper and lower bounds of the preset target ant’s dimension values. and and represent the upper and lower bounds of the ant’s search range for each dimension in the tth iteration, respectively. The value of R is determined according to the current number of iterations.

[0089] R can be determined by the following formula:

[0090] The value of w depends on the current iteration number t, T represents the maximum number of iterations, When w=0; When w=2; Then w=3; Then w=4; when When w=5; when When w=6, so that It shows a segmented exponential increasing trend.

[0091] The walking step size determines the movement range of the ant at each position update. With each iteration, the walking step size of the ant will change with the change of the random number. The step size r is calculated based on the generated random number. In each round of iteration, the step size r will affect the current position X(t). Specifically, the current position X(t) is the cumulative sum of the step sizes generated in all iterations, indicating the overall change from the initial position to the current iteration position, that is, the current position of the ant is obtained by adding the random walking step sizes corresponding to each round of iteration.

[0092] The current position of the target ant can be determined by the following formula:

[0093] Where t is the current iteration number, T is the maximum iteration number, X(t) represents the random walk position, cumsum represents the cumulative sum of the random walk steps in the previous t-1th iteration, and r is the random walk step generation function, which is used to calculate the walk step corresponding to each iteration. The calculation is: , where rand represents a random number between (0,1).

[0094] By adopting the embodiments of the present disclosure, the wandering range is reduced as the number of iterations increases, so that the wandering range is larger in the initial stage, which is conducive to extensive exploration. As the number of iterations increases, the wandering range gradually decreases, helping ants to perform local refinement optimization after finding a better solution.

[0095] In this round of iteration, the new position vector of the target ant will be encoded according to the second encoding format to obtain a continuous encoding group of the target ant. The continuous encoding group usually includes the updated parameters of the target ant, such as thresholds and weights.

[0096] After the discrete code group of the target ant is determined by ACO and the continuous code group of the target ant is determined by ALO, the position code of the target ant is obtained according to the discrete code group and the continuous code group.

[0097] By adopting the embodiment of the present disclosure, the feature selection method based on elite ant guidance can improve the efficiency of the search process. The target ants can inherit the excellent feature selection strategy of the elite ants by imitating their selection, thereby avoiding blindly searching the feature space from the beginning. By adjusting the wandering range and step length, the target ants can flexibly adjust the search direction and pace in the feature space, effectively avoiding falling into the local optimal solution or premature convergence. The guiding role of the elite ants makes the search of the target ants more directional, and can maintain the ability of global optimization in a larger search space, thereby improving the global search ability and optimization efficiency of the algorithm. The combination of position coding provides a comprehensive and accurate representation method, and the isolated forest corresponding to the target ant is fully described. This coding form enables each ant to flexibly search and adjust in the feature space, improving the efficiency of the algorithm and the search quality.

[0098] Among them, in an optional embodiment, according to the elite ants obtained in the previous round of iteration, the target ants select feature subsets corresponding to each isolated tree in the isolated forest from the feature selection space, including: determining the pheromone concentration of each feature in the feature selection space in this round of iteration according to the discrete coding group corresponding to the elite ants obtained in the previous round of iteration; determining the number of selection steps of the target ant, and the number of selection steps is used to determine the number of features included in the feature subset; when the target ant selects features from the optional features in the feature search space at each step, determining the selection probability of each optional feature according to the pheromone concentration and heuristic information of each optional feature; the optional feature represents the unselected feature; based on the selection probability of the optional feature, select an optional feature as the feature corresponding to the step according to the roulette strategy; determine the feature subset of the isolated tree according to the features corresponding to each step; after determining the feature subset of an isolated tree, clear the selected state of the feature in the feature selection space.

[0099] In the ACO algorithm, ants pass experience to each other through pheromones to influence the decisions of other ants. Elite ants are the ants with the highest fitness determined in the previous iteration. Therefore, the pheromone concentration of each feature in the feature selection space can be updated through the discrete coding group corresponding to the elite ants, so that the elite ants determined in the previous iteration can guide the M ants in this iteration to determine their corresponding position coding.

[0100] Among them, in an optional embodiment, according to the discrete coding group corresponding to the elite ant obtained in the previous round of iteration, the pheromone concentration of each feature in the feature selection space in the current round of iteration is determined, including: according to the discrete coding group corresponding to the elite ant obtained in the previous round of iteration, the feature subset of the elite ant is determined; the features of the feature subset corresponding to the elite ant are subjected to pheromone enhancement processing, and the features of the feature subset corresponding to the remaining ants obtained in the previous round of iteration are subjected to pheromone volatilization processing; through the pheromone enhancement processing and the pheromone volatilization processing, the pheromone concentration of each feature in the feature selection space of the current round of iteration is determined.

[0101] For the feature subsets corresponding to the elite ants, it is necessary to enhance these features with pheromones, that is, to increase the weight of the features selected by the elite ants, so that these features are more likely to be selected in the next iteration. By enhancing pheromones, more search resources can be concentrated on those features that perform well, thereby accelerating the process of finding the optimal feature subset.

[0102] The opposite of pheromone reinforcement is pheromone volatilization, which means reducing the weight of features selected by ants with poor performance. Pheromone volatilization makes those features that fail to perform well gradually lose their appeal, thereby encouraging ants to select new feature subsets and maintain the diversity of the search space.

[0103] The pheromone concentration of each feature in the feature selection space is obtained by processing the selection results of the elite ants and other ants obtained in the previous iteration. Through the pheromone enhancement processing of the elite ant feature subset and the pheromone volatilization processing of the other ant feature subsets, the updated pheromone concentration of all features in this iteration is finally determined. These pheromone concentration values ​​will affect the decision of M ants in the next round of feature selection.

[0104] The pheromone of each feature in the feature selection space can be updated by the following formula:

[0105] in: It indicates the volatility coefficient of pheromone; Indicates the path solution (feature subset) that needs to be updated; represents the evaluation value of the path solution on the objective function; Q is a constant, which is the scaling factor of the objective function evaluation value. The factor value is determined by the initial value of the pheromone, the volatility coefficient and the objective function evaluation value. Among them, the objective function is pre-set and can be used to measure the performance of elite ants on a certain task.

[0106] By using the embodiments of the present disclosure, the selection result of the elite ants is usually the solution with the best quality in the current search space. By enhancing the pheromone concentration of these features, it is possible to accelerate the convergence toward the optimal feature subset, and it is possible to find a relatively excellent feature subset in fewer iterations, avoiding the need for a large number of invalid searches. This balance allows the algorithm to ensure local search optimization while not losing the ability of global search, thereby avoiding falling into the local optimal solution and being able to find the global optimal solution.

[0107] The number of selection steps of the target ant determines how many features the target ant needs to select when selecting features for a certain isolated tree. The number of selection steps is usually set according to the requirements of the problem or the maximum number of features, ensuring that the size of the feature subset is moderate and avoiding too many or too few feature selections.

[0108] Optional features represent features that are not selected in the feature selection space, that is, the features included in the feature subset are not repeated. Moreover, the selection of features in each feature subset is selected from the complete feature selection space. For example, for a feature subset of an isolated tree, when the first feature included in the feature subset is selected from the feature selection space, none of the features in the feature selection space have been selected. After the feature subset of an isolated tree is determined, the selection status of each feature in the feature selection space can be cleared. When selecting features from the feature selection space for the feature subset of the next isolated tree, feature selection is performed again from the complete feature selection space.

[0109] Figure 3 FIG. 1 is a schematic diagram of a feature selection process shown in an embodiment of the present disclosure. Figure 3 As shown in the figure, f represents the feature, the feature selection space includes n features, s represents the number of steps selected by the ant, and q represents the maximum number of steps selected by the ant. Initially, the selected features in the feature selection space are empty, and the ant starts selecting features from the first step and stops at the qth step. Each time, an unvisited feature is selected from the optional features according to the roulette strategy.

[0110] Each time the target ant selects a feature from the optional features, the selection probability is calculated based on a comprehensive consideration of the pheromone concentration and heuristic information. The selection probability of the jth feature in the feature selection space can be determined by the following formula:

[0111] in, represents the pheromone value of the jth feature; Represents the heuristic information of the jth feature; Indicates the set of features that ant a has visited; and Represents the influence factor of pheromone and heuristic information.

[0112] Among them, the heuristic information uses the information gain of the feature, which measures the reduction in uncertainty before and after the feature divides the data set. The larger the information gain, the stronger the relevance of the feature. Uncertainty can be represented by entropy. Information gain is an indicator to measure the importance of a feature in data division. The larger the information gain of a feature, the more effective it is in reducing the uncertainty of the data, and therefore the stronger the relevance.

[0113] The target ant uses a roulette wheel strategy to select a feature based on the probability of selecting the optional features. In this process, pheromone concentration and heuristic information work together to ensure that the selected features are the most important for building an isolated tree. The target ant builds a feature subset of an isolated tree based on the features selected at each step. When determining the feature subset corresponding to the next isolated tree, it reselects features from the feature selection space based on the number of selection steps.

[0114] By adopting the embodiments of the present disclosure, according to the distribution of pheromone concentration, the target ants can be more inclined to select those features that performed well in the previous round of iterations when selecting features, and can effectively guide the feature selection to efficient and important features, avoiding the selection of irrelevant features, thereby accelerating the search process and improving the accuracy of the selection. The roulette strategy maintains the diversity of the search by introducing randomness and avoids falling into the local optimal solution. By setting the number of selection steps of the target ants, the number of features used by each isolated tree can be accurately controlled, thereby ensuring the quality of each isolated tree in the isolated forest, helping to improve the effect of the isolated forest model and avoiding overfitting or underfitting. By ensuring that the features in each feature subset are selected from the feature selection space and are not repeated, the addition of redundant features can be avoided, the independence and diversity of each feature selection can be guaranteed, and the feature subset of each isolated tree is not affected by the selection results of other isolated trees, avoiding the overlap of features between different trees or the selection bias towards specific features.

[0115] Among them, in an optional embodiment, when the target ant selects a feature from the optional features in the feature search space at each step, the selection probability of each optional feature is determined according to the pheromone value and heuristic information of each optional feature, including: determining a multiple selection coefficient of the optional feature, the multiple selection coefficient is used to control the diversity in the feature selection process; the value corresponding to the multiple selection coefficient of the feature is determined according to the number of times the feature is selected in this round of iterations; the selection probability of each optional feature is determined according to the multiple selection coefficient, pheromone value and heuristic information of the optional feature.

[0116] In order to increase the diversity of isolated trees, multiple selection coefficients can be introduced. Diversity refers to the differences in the features selected by ants during the search process. Improving the diversity of isolated trees can avoid always selecting similar feature subsets for isolated trees. The multiple selection coefficient is used to control the probability of each feature being selected multiple times during the feature selection process. Specifically, if a feature is selected multiple times, its multiple selection coefficient will increase, which means that the feature is less likely to be selected in subsequent selections.

[0117] The multiple selection coefficient of a specific feature is calculated dynamically and is related to the number of times the specific feature is selected in the current iteration. Specifically, in each round of iteration, the number of times each feature is selected is recorded, and the multiple selection coefficient of each feature is calculated based on the number of times.

[0118] After introducing multiple selection coefficients, the selection probability of optional features in the feature selection space can be determined by the following formula:

[0119] in, represents the pheromone value of the jth feature; Represents the heuristic information of the jth feature; Indicates the set of features that ant a has visited; and represents the influence factor of pheromone and heuristic information, is the multiple selection coefficient of the cth isolated tree, represents the number of times the jth feature was selected by the first c-1 isolated trees, for example, Indicates the number of times the 8th feature was selected by the first 10 isolated trees.

[0120] By adopting the embodiments of the present disclosure and introducing multiple selection coefficients, it is possible to effectively avoid over-reliance on certain features during the search process and ignore other possible important features, thereby ensuring that the ants are sufficiently exploratory in feature selection and avoiding the dilemma of local optimal solutions. The multiple selection coefficients help the algorithm maintain sufficient diversity during the search process, while strengthening the selection of high-quality features through pheromones and heuristic information, so that development and exploration are balanced, so that the search process will not converge to the local optimal solution too early, and the search strategy can be continuously adjusted to explore new feature combinations.

[0121] Among them, in an optional embodiment, the target ant updates the position vector of the target ant according to the walking range and walking step length of this round of iteration and based on the elite ant obtained in the previous round of iteration, including: selecting a roulette ant from the M second ants in the previous round of iteration in a roulette manner; according to the walking range and walking step length of this round of iteration, the target ant respectively surrounds the elite ant and the roulette ant obtained in the previous round of iteration to determine the elite position vector and the roulette position vector; according to the elite position vector and the roulette position vector, the position vector of the target ant is updated.

[0122] The roulette ant is an ant selected from the M ants input in this round (i.e., the M second ants in the previous iteration) based on a roulette method. The roulette ant represents an ant that performed well in the previous iteration. The roulette ant plays a certain guiding role in the update of the target ant.

[0123] When updating the position vector of the target ant in this iteration around the elite ant, it is also necessary to synchronously determine a position vector around the roulette ant. The position vector obtained around the elite ant is determined as the elite position vector, and the position vector determined around the roulette ant is determined as the roulette position vector.

[0124] When walking around the elite ant and the roulette ant respectively, the walking step length and the walking range are consistent, and the same walking step length and walking range are used to determine the position vector around the elite ant and the roulette ant respectively.

[0125] After obtaining the elite position vector and the roulette position vector, the position vector of the target ant in this iteration is updated by combining the elite position vector and the roulette position vector. The position vector is used to obtain the continuous coding part in the position coding of the target ant in this iteration.

[0126] Specifically, the position vector of the target ant in this iteration can be determined in the following way:

[0127] Among them, At is the position vector of the ant in the tth iteration, e and r represent elite ants and roulette ants respectively. is the elite position vector obtained by the target ant's random walk around e in the tth iteration, The roulette vector obtained by the target ant's random walk around r in the tth iteration.

[0128] By adopting the embodiments of the present disclosure, by considering the influence of elite ants and roulette ants simultaneously in position updating, the target ants can explore in multiple directions, not just in the area of ​​elite ants. The guidance of roulette ants enables the ants to maintain appropriate diversity around the current solution, avoiding the over-concentration of the search process and reducing the risk of falling into a local optimum. Considering the influence of elite ants and roulette ants at the same time allows the ants to move closer to the optimal solution while not ignoring other potential high-quality solutions, thereby enhancing the global optimization capability of the algorithm and being able to better find the global optimal solution in a large-scale or high-dimensional search space.

[0129] In order to implement the high-dimensional abnormal data detection method provided by the present disclosure, the present disclosure proposes a high-dimensional abnormal data detection algorithm. Figure 4 It is a schematic diagram of the steps of a high-dimensional abnormal data detection algorithm shown in an embodiment of the present disclosure.

[0130] according to Figure 4 As shown in the figure, the input of the algorithm is the maximum number of iterations T, the objective function F for updating pheromones, and the number of ants M. After T rounds of iterations, the output is an anomaly detection model for high-dimensional data detection. M ants means that in each round of iteration, M ants will update the position code respectively.

[0131] Step S1: Initialize the position codes of M ants.

[0132] The position coding of the ants is updated in each round of iteration, and a corresponding isolated forest can be obtained based on the position coding of the ants. The M ants with initialized position coding are used as the M ants input in the first round of iteration, and their initialized position coding represents the position coding before the first round of iteration update.

[0133] Step S2: Calculate the fitness values ​​of all ants and select the individuals with the best fitness values ​​from the population as elite ants.

[0134] The fitness value of the initialized ant is calculated, and the fitness value is used to judge the effect of the isolated forest corresponding to the ant in detecting abnormal data. The elite ants selected before entering the iteration process are the elite ants that provide guidance in the first round of iteration.

[0135] The iterative process begins from step S3.

[0136] Step S3: Determine whether the current number of iterations t is less than the maximum number of iterations T.

[0137] If the judgment is yes, the process proceeds to step S4 ; if the judgment is no, the process proceeds to step S10 .

[0138] Step S4: Determine whether the current mth ant is less than the maximum number of ants M.

[0139] If the judgment is yes, proceed to step S5; if the judgment is no, proceed to step S8.

[0140] Step S5: Determine whether the current i-th isolated tree is less than the maximum number of isolated trees I.

[0141] If the judgment is yes, proceed to step S6; if the judgment is no, proceed to step S7.

[0142] Step S6: The mth ant calculates the path probability value that meets the conditions at a certain step for the i-th isolated tree, selects a path using the roulette strategy, and finally generates a path solution for the isolated tree. The path solution represents the feature subset corresponding to the current isolated tree i.

[0143] The mth ant determines the feature subset corresponding to the ith isolated tree. The feature subset includes q features, and m ants make q steps of selection. In any step of selection, the probability value of the path that meets the conditions is calculated, that is, the selection probability of the optional features in the feature selection space is calculated; then a roulette strategy is used to select a certain path, that is, a feature, and finally a path solution is generated for the ith isolated tree. The path solution represents the feature subset of the isolated tree.

[0144] The mth ant continues to determine the feature subset corresponding to the i+1th isolated tree, returns to step S5, and continues to determine whether the i=i+1th isolated tree is less than the maximum number of isolated trees I.

[0145] After I rounds of step S6, the mth ant completes the determination of feature subsets for I isolated trees, encodes the features in the I feature subsets, and obtains a discrete encoding group for the mth ant.

[0146] Step S7: A roulette ant is selected through roulette, and the ants update the continuous coding group around the elite ants and the roulette ants.

[0147] A roulette ant is selected from M ants through the roulette strategy, and ant m walks around the roulette ant and the elite ant respectively, thereby updating the continuous coding group in the position coding of the mth ant.

[0148] After steps S5 to S7, the position code of the mth ant is obtained, and the position code of the mth ant is updated to obtain the mth first ant of this iteration. The position code of the m+1th ant is continued to be determined, and the process returns to step S4 to continue to determine whether the m=m+1th ant is less than the maximum number of ants M.

[0149] Step S8: Decode the position codes of the M ants obtained in the tth iteration respectively, obtain the corresponding isolated forest and evaluate it, obtain the fitness values ​​of the M ants, apply the matching exchange strategy to generate new ants and evaluate it, and obtain the fitness value of the new ants.

[0150] The position codes of the M ants obtained in the t-th iteration are updated position codes, and the M ants after the updated position codes in the t-th iteration represent the M first ants in the t-th iteration.

[0151] After decoding the position codes of the M first ants in the t-th iteration, the corresponding M isolated forests are obtained, and the fitness values ​​of the M first ants in the t-th iteration are obtained based on the abnormal data detection effect of the isolated forests. Based on the fitness values ​​of the M first ants in the t-th iteration, the M first ants in the t-th iteration are divided into a better ant pool and a worse ant pool. The matching exchange strategy is applied to exchange the continuous codes of the better ants and the worse ants in the matching ant group to obtain new ants. The fitness value of the new ants is also determined.

[0152] Step S9: Update elite ants based on the fitness values ​​of the ants, and use the elite ants to update pheromones.

[0153] Based on the fitness value of the new ant, the fitness values ​​of the M first ants in the tth iteration, and the M ants before the position code update in the tth iteration (the M second ants in the previous iteration), the M ants input to the next round are updated to obtain the updated M ants (i.e., the M second ants in this iteration). Among them, the position codes of the M second ants in this iteration are updated in the t+1th iteration.

[0154] Based on the updated fitness values ​​of the M ants, the elite ants are updated and the pheromones are updated using the elite ants. The updated pheromones and elite ants are used to determine the position codes of the M ants in the t+1th iteration.

[0155] Enter the t+1th iteration, return to step S3, and continue to determine whether t=t+1 is less than the maximum number of iterations T.

[0156] Step S10: Generate an anomaly detection model based on the elite ants determined in the last round of iteration.

[0157] The position code of the elite ants obtained in the last round of iteration is decoded to obtain the corresponding isolation forest, and an anomaly detection model is generated based on the isolation forest.

[0158] Based on the same technical concept, the present disclosure provides a high-dimensional abnormal data detection device. Figure 5 Schematic diagram of a high-dimensional abnormal data detection device shown in an embodiment of the present disclosure. Figure 5 As shown, the device comprises: A determination module 510 is used to determine the position codes of the M first ants in this iteration, where the position code of each ant includes: a discrete code group for determining an isolated tree feature subset, and a continuous code group for determining an isolated tree abnormality threshold and an isolated tree weight, where each feature in the isolated tree feature subset represents a segmentation method for a high-dimensional data set; A construction module 520 is used to construct M isolated forests according to the position codes of the M first ants in the current iteration; An evaluation module 530 is used to evaluate the effectiveness of the M isolation forests based on high-dimensional abnormal data detection, and determine the fitness value corresponding to each of the M first ants in the current iteration, wherein the fitness value represents the detection effect of abnormal data in the high-dimensional data set; An updating module 540 is used to determine an elite ant in this iteration according to the fitness values ​​corresponding to the M first ants in this iteration and the fitness values ​​corresponding to the M second ants in the previous iteration, wherein the elite ant is used to determine the position codes of the M first ants in the next iteration; Iteration module 550 is used to perform multiple iterations according to the above steps, and determine the final target position code according to the position code corresponding to the elite ant determined in the last round of iteration, and the target position code includes: optimal feature subset selection, isolated tree abnormality threshold and isolated tree weight; End module 560 is used to build an anomaly detection model according to the final target position code, and segment the high-dimensional data set based on the data segmentation method corresponding to the anomaly detection model to obtain an anomaly detection result.

[0159] The present disclosure also provides an electronic device, referring to Figure 6 , Figure 6 is a schematic diagram of an electronic device shown in an embodiment of the present disclosure. Figure 6As shown, the electronic device 600 includes: a memory 610 and a processor 620. The memory 610 and the processor 620 are connected via a bus communication. A computer program is stored in the memory 610. The computer program can be run on the processor 620 to implement the steps in the high-dimensional abnormal data detection method disclosed in the embodiment of the present disclosure.

[0160] The embodiment of the present disclosure further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the high-dimensional abnormal data detection method disclosed in the embodiment of the present disclosure are implemented.

[0161] The embodiment of the present disclosure further provides a computer program product, including a computer program. When the computer program is executed by a processor, the steps in the high-dimensional abnormal data detection method disclosed in the embodiment of the present disclosure are implemented.

[0162] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0163] It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, devices or computer program products. Therefore, the embodiments of the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0164] The embodiments of the present disclosure are described with reference to the flowcharts and / or block diagrams of the methods, apparatuses, electronic devices, and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0165] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0166] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0167] Although some embodiments of the present disclosure have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the present disclosure.

[0168] The above is a detailed introduction to a high-dimensional abnormal data detection method provided by the present disclosure. In this article, specific examples are used to illustrate the principles and implementation methods of the present disclosure. The description of the above embodiments is only used to help understand the method of the present disclosure and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present disclosure, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present disclosure.

Claims

1. A high-dimensional abnormal data detection method, characterized in that: include: Determine the position codes of the M first ants in this round of iteration, where the position code of each ant includes: a discrete code group for determining an isolated tree feature subset, and a continuous code group for determining an isolated tree abnormality threshold and an isolated tree weight, where each feature in the isolated tree feature subset represents a segmentation method for a high-dimensional data set; Constructing M isolation forests according to the position codes of the M first ants in the current iteration; Based on high-dimensional abnormal data detection, the effectiveness of the M isolation forests is evaluated respectively, and the fitness value corresponding to each of the M first ants in the current iteration is determined, and the fitness value represents the detection effect of abnormal data in the high-dimensional data set; Determine the elite ant of this iteration according to the fitness values ​​corresponding to the M first ants of this iteration and the fitness values ​​corresponding to the M second ants of the previous iteration, wherein the elite ant is used to determine the position codes of the M first ants of the next iteration; Perform multiple iterations according to the above steps, and determine the final target position code based on the position code corresponding to the elite ant determined in the last round of iteration. The target position code includes: optimal feature subset selection, isolated tree anomaly threshold and isolated tree weight; According to the final target position code, an anomaly detection model is constructed, and based on the data segmentation method corresponding to the anomaly detection model, the high-dimensional data set is segmented to obtain an anomaly detection result.

2. The method according to claim 1, characterized in that Determining the elite ant of the current iteration according to the fitness values ​​corresponding to the M first ants of the current iteration and the fitness values ​​corresponding to the M second ants of the previous iteration comprises: Sorting the M first ants in the current iteration and the M second ants in the previous iteration according to the fitness values ​​corresponding to the M first ants in the current iteration and the fitness values ​​corresponding to the M second ants in the previous iteration to obtain a first sorting result; Determine the first M ants in the first sorting result as the M second ants in this round of iteration; The ant with the highest fitness among the M second ants in the current iteration is determined as the elite ant in the current iteration.

3. The method according to claim 2, characterized in that Also includes: Sorting the M first ants in the current iteration according to the fitness values ​​corresponding to the M first ants in the current iteration to obtain a second sorting result; Taking half of the ants with high fitness values ​​in the second sorting result corresponding to the M first ants in the current iteration as a better ant pool, and taking half of the ants with low fitness values ​​in the second sorting result as a worse ant pool; According to a preset matching number, ants are selected from the better ant pool and the worse ant pool respectively to obtain a matching number of matching ant groups; each matching ant group includes an ant from the better ant pool and an ant from the worse ant pool; For each matching ant group, the two ants included in the group exchange their continuous codes to obtain new ants; According to the fitness values ​​corresponding to the M first ants of the current iteration and the fitness values ​​corresponding to the M second ants of the previous iteration, the M first ants of the current iteration and the M second ants of the previous iteration are sorted to obtain a first sorting result, including: The new ants, the M first ants in the current iteration, and the M second ants in the previous iteration are sorted according to their respective corresponding fitness values ​​to obtain a first sorting result.

4. The method according to claim 1, characterized in that: According to the position codes of the M first ants in the current iteration, M isolation forests are constructed, including: according to the position code of each first ant, an isolation forest is constructed; Each isolation forest is constructed as follows: Decoding each discrete code in the discrete code group that is sorted in sequence to obtain each feature that is sorted in sequence; According to the order of each feature, determine the feature subset of each isolated tree in the isolated forest corresponding to the first ant; Decoding the first continuous code in the continuous code group to obtain the isolated tree abnormality threshold of the first ant; Decoding the remaining sequentially ordered continuous codes in the sequentially ordered group to obtain respective weight values ​​that are sequentially ordered; Determine the weight of each isolated tree in the isolated forest corresponding to the first ant according to the order of each weight value; An isolated forest corresponding to the first ant is constructed according to the feature subsets corresponding to the isolated trees, the weights corresponding to the isolated trees, and the isolated tree abnormality threshold.

5. The method according to claim 1, characterized in that Determine the position codes of the M first ants in this round of iteration, including: According to the elite ants obtained in the previous round of iteration, the target ant selects feature subsets corresponding to each isolated tree in the isolated forest from the feature selection space, wherein the feature subsets include multiple features; wherein the feature subsets are sorted in order, and each feature in the feature subsets is sorted in order, and the target ant is any one of the M second ants in the previous round of iteration; Encoding the features in the feature subset according to a first encoding format to obtain discrete codes of the features in the discrete code group; The target ant updates the position vector of the target ant according to the wandering range and wandering step length of this iteration and based on the elite ant obtained in the previous iteration; According to the position vector of the target ant, encoding is performed according to a second encoding format to obtain a continuous encoding group corresponding to the target ant; The position code of any one of the M first ants in the current iteration is determined according to the discrete code group composed of the discrete codes and the continuous code group.

6. The method according to claim 5, characterized in that According to the elite ants obtained in the previous iteration, the target ants select feature subsets corresponding to each isolated tree in the isolation forest from the feature selection space, including: According to the discrete coding group corresponding to the elite ants obtained in the previous iteration, the pheromone concentration of each feature in the feature selection space in this iteration is determined; Determine the number of selection steps of the target ant, where the number of selection steps is used to determine the number of features included in the feature subset; In the case where the target ant selects a feature from the optional features in the feature search space at each step, determining the selection probability of each optional feature according to the pheromone concentration and heuristic information of each optional feature; the optional feature represents the feature that is not selected; Based on the selection probability of the optional feature, select an optional feature as the feature corresponding to the step according to the roulette strategy; According to the features corresponding to each step, determine the feature subset of the isolated tree; After determining a feature subset of an isolated tree, the selected state of the features in the feature selection space is cleared.

7. The method according to claim 6, characterized in that In the case where the target ant selects a feature from the optional features in the feature search space at each step, determining the selection probability of each optional feature according to the pheromone value and heuristic information of each optional feature, including: Determine a multiple selection coefficient of the optional feature, wherein the multiple selection coefficient is used to control the diversity in the feature selection process; a value corresponding to the multiple selection coefficient of the feature is determined according to the number of times the feature is selected in this round of iteration; The selection probability of each optional feature is determined according to the multiple selection coefficients, pheromone values ​​and heuristic information of the optional features.

8. The method according to claim 6, characterized in that According to the discrete coding group corresponding to the elite ants obtained in the previous iteration, the pheromone concentration of each feature in the feature selection space in this iteration is determined, including: Determine a feature subset of the elite ant according to the discrete coding group corresponding to the elite ant obtained in the previous iteration; Performing pheromone enhancement processing on the features of the feature subset corresponding to the elite ants, and performing pheromone volatilization processing on the features of the feature subsets corresponding to the remaining ants obtained in the previous round of iterations; The pheromone concentration of each feature in the feature selection space of this round of iteration is determined through the pheromone enhancement process and the pheromone volatilization process.

9. The method according to claim 5, characterized in that The target ant updates the position vector of the target ant according to the wandering range and wandering step length of this iteration based on the elite ant obtained in the previous iteration, including: Selecting a roulette ant from the M second ants of the previous iteration in a roulette manner; According to the wandering range and wandering step length of this round of iteration, the target ant respectively surrounds the elite ant and the roulette ant obtained in the previous round of iteration to determine the elite position vector and the roulette position vector; The position vector of the target ant is updated according to the elite position vector and the roulette position vector.

10. The method according to claim 5 or 9, characterized in that: The walking step length and the walking range are determined by the following steps: Determine preset upper and lower bounds of the dimension of the target ant; Determine the wandering range of the target ant according to the upper and lower bounds of the dimension of the target ant, the current iteration number corresponding to this round of iteration, and the maximum iteration number; According to the current iteration number corresponding to the current iteration, the walking step length of the target ant in the current iteration is determined.

Images