Balanced classification method of radar operation data based on SMOTE

Abstract

The embodiments of the present application relate to the technical field of radar operation data analysis, and in particular to a balanced classification method for radar operation data based on SMOTE, including: classifying radar operation data, dividing it into majority class samples and minority class samples according to normal operation data and fault operation data; dividing each minority class sample into minority class boundary samples and minority class internal samples based on the original distribution of radar operation data; processing minority class boundary samples using K-L-SMOTE algorithm to obtain first class synthetic samples; dividing minority class internal samples into different subclusters using K-means clustering algorithm, calculating the sampling multiple and synthesizing them to obtain second class synthetic samples; merging the first class synthetic samples, the second class synthetic samples and the majority class samples to obtain a balanced sample set; training the classifier based on the balanced sample set, and performing radar operation status analysis based on the classification results. This method can well avoid the problem of imbalanced minority class samples.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of radar operation data analysis, and in particular to a radar operation data balance classification method based on SMOTE. Background Art

[0002] The faulty operation data generated by the radar during operation is far less than the normal operation data. If the potential faulty operation data is mistakenly classified as normal operation data, it is likely to cause a major communication accident and cause inestimable property losses. Therefore, in actual situations, it is very important to classify the minority class data (i.e., faulty operation data) more accurately. Traditional classification learning methods often favor the majority class and ignore the minority class. The misclassification cost of the minority class is usually much higher than the classification cost of the majority class, which poses a huge safety hazard.

[0003] At present, there are two main methods for processing radar operation data sets: random oversampling and random undersampling. Among the random oversampling techniques, SMOTE (Synthetic Minority Oversampling Technique) is the most popular method. Although SMOTE has been proven to be a simple and effective oversampling method, it also has some disadvantages, such as unclear separation between the majority class group and the minority class group, and easy to generate noise. Therefore, technicians usually select minority class samples to synthesize as new samples, but this method ignores the surrounding environment and brings two new problems. The first problem is that if the surrounding distribution of the selected minority class samples is all minority class samples, then the newly synthesized samples are of little significance, because these newly synthesized samples are often far away from the boundary between the majority class and the minority class. The second problem is that when generating minority class samples through SMOTE, it is necessary to randomly select k nearest neighbors for interpolation, which does not take into account the uneven distribution of minority class samples. In other words, the newly synthesized samples by random interpolation are likely to change the distribution of the original minority class samples. Summary of the invention

[0004] In order to solve the above technical problems, the embodiment of the present application proposes a balanced classification method for radar operation data based on SMOTE, which can perform targeted processing on samples at the boundary between the majority class and the minority class and samples within the minority class, thereby effectively avoiding the problem of imbalance of minority class samples, and further effectively improving the analysis and processing accuracy of radar operation data, and reducing the probability of communication accidents.

[0005] In order to achieve the above-mentioned purpose, an embodiment of the present application proposes a balanced classification method for radar operation data based on SMOTE, comprising the following steps: performing classification preprocessing on the acquired radar operation data, and dividing it into majority class samples and minority class samples according to normal operation data and fault operation data; based on the original distribution of the radar operation data, dividing each minority class sample into minority class boundary samples and minority class internal samples; wherein the minority class boundary samples are minority class samples distributed at the classification boundary between the majority class and the minority class, and the minority class internal samples are minority class samples distributed inside the minority class; processing the minority class boundary samples using the KL-SMOTE algorithm to obtain a first class of synthetic samples; dividing the minority class internal samples into different subclusters using the K-means clustering algorithm, and calculating different sampling multiples according to the number of minority class internal samples in each subcluster for synthesis to obtain a second class of synthetic samples; merging the first class of synthetic samples, the second class of synthetic samples and the majority class samples to obtain a balanced sample set; based on the balanced sample set, training and testing different classifiers, and performing radar operation status analysis based on the classification results of each classifier.

[0006] In order to achieve the above-mentioned purpose, an embodiment of the present application further proposes a balanced classification system for radar operation data based on SMOTE, and the system includes: a majority-minority division module, a boundary internal division module, a first category synthesis module, a second category synthesis module, a merging module and an analysis and processing module; the majority-minority division module is used to perform classification preprocessing on the acquired radar operation data, and divide it into majority class samples and minority class samples according to normal operation data and fault operation data; the boundary internal division module is used to divide each minority class sample into minority class boundary samples and minority class internal samples based on the original distribution of the radar operation data, wherein the minority class boundary samples are minority class samples distributed on the classification boundary between the majority class and the minority class, and the minority class internal samples are The samples are minority class samples distributed inside the minority class; the first class synthesis module is used to process the minority class boundary samples using the KL-SMOTE algorithm to obtain the first class synthetic samples; the second class synthesis module is used to divide the minority class internal samples into different subclusters using the K-means clustering algorithm, and calculate different sampling multiples according to the number of minority class internal samples in each subcluster for synthesis to obtain the second class synthetic samples; the merging module is used to merge the first class synthetic samples, the second class synthetic samples and the majority class samples to obtain a balanced sample set; the analysis and processing module is used to train and test different classifiers based on the balanced sample set, and perform radar operation status analysis based on the classification results of each classifier.

[0007] In order to achieve the above-mentioned purpose, an embodiment of the present application also proposes an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a radar operation data balancing classification method based on SMOTE as described above.

[0008] In order to achieve the above objectives, an embodiment of the present application proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement a radar operation data balance classification method based on SMOTE as described above.

[0009] The radar operation data balanced classification method based on SMOTE proposed in the embodiment of the present application first divides the acquired radar operation data into majority class and minority class according to the normal operation data and the fault operation data, and then divides the boundary and internal of each minority class sample based on the original distribution of the radar operation data. For the minority class boundary samples, the present application chooses to use the KL-SMOTE algorithm for processing to obtain the first class of synthetic samples to expand the minority class boundary samples. For the minority class internal samples, the present application chooses to use the K-means clustering algorithm to divide the sub-clusters, and then calculates different sampling multiples according to the number of minority class internal samples in each sub-cluster for synthesis, and obtains the second class of synthetic samples, thereby achieving intra-class balance between minority class internal samples. Finally, the first class of synthetic samples, the second class of synthetic samples and the majority class samples are merged to obtain a balanced sample set to train and test the classifier, and then the radar operation status analysis is performed based on the classification results. After targeted processing of minority class boundary samples and minority class internal samples, the problem of imbalance of minority class samples is effectively avoided, thereby improving the analysis and processing accuracy of radar operation data and reducing the probability of traffic accidents.

[0010] In some optional embodiments, the minority class samples in the radar operation data are recorded as c i , i = 1, 2, ..., g, g is the total number of minority class samples. Based on the original distribution of radar operation data, each minority class sample is divided into minority class boundary samples and minority class internal samples, including: for minority class samples c i , calculate c by Euclidean distance i of Neighbor, record this The number of samples belonging to the majority class among the nearest neighbors And compare and The size of Then c iDivided into minority class boundary samples c ib , and stored in the set B{c ib};like Then c i Divided into minority class internal samples c ii , and stored in the set I{c ii};like Then c i are considered as noise samples and removed.

[0011] In some optional embodiments, the minority class boundary samples are processed using the KL-SMOTE algorithm to obtain the first type of synthetic samples, including: for the minority class boundary samples c ib , calculate c ib In the set B{c ib}, and are denoted by c iμ ; Based on c ib and c iμ Interpolate to generate the first type of synthetic samples and will Save to collection The interpolation formula is specifically expressed as random(0,1) means generating a random number between 0 and 1. That is, the collection of the first type of synthetic samples.

[0012] In some optional embodiments, the method of dividing the minority class internal samples into different sub-clusters using the K-means clustering algorithm, and calculating different sampling multiples according to the number of minority class internal samples in each sub-cluster for synthesis to obtain the second type of synthetic samples includes: for the set I{c ii}, the K-means clustering algorithm is used to divide it into n different sub-clusters and calculate the cluster center of each sub-cluster, denoted as {β 1 , β 2 , ..., β j , ..., β n}, β j is the cluster center of the jth subcluster;

[0013] The sampling rate N of each sub-cluster is calculated based on the number of samples within the minority class in each sub-cluster. j , sampling rate N j The calculation formula is expressed as:

[0014]

[0015] in, represents the number of samples within the minority class that accounts for the largest proportion in the j-th subcluster, represents the number of samples in the minority class with the smallest proportion in the j-th subcluster, and size(·) represents the size function;

[0016] For the jth sub-cluster, based on the minority class internal samples and β j Synthesize the second type of synthetic samples to obtain the second type of synthetic samples Among them, the number of the second type of synthetic samples that need to be synthesized is N j The product of the number of samples in the minority class in the jth subcluster is expressed as

[0017] In some optional embodiments, the value of n ranges from 27 to 34.

[0018] In some optional embodiments, the radar operation data includes normal operation data, slight fault operation data, general fault operation data, and severe fault operation data, the normal operation data includes normal operation data, and the fault operation data includes slight fault operation data, general fault operation data, and severe fault operation data;

[0019] Alternatively, the normal operation data includes ordinary operation data and minor fault operation data, and the fault operation data includes general fault operation data and major fault operation data.

[0020] In some optional embodiments, different classifiers include GBDT, KNN and random forest; different classifiers are trained and tested based on a balanced sample set, including: based on a test set and a balanced sample set, different classifiers are trained and tested; wherein the test set is obtained by dividing from the radar operation data by a K-fold cross-validation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related technologies, the drawings required for use in the embodiments of the present application or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 is a flowchart of a radar operation data balance classification method based on SMOTE provided in one embodiment of the present application;

[0023] Figure 2 is a schematic diagram of G-mean experimental results using a traditional SMOTE oversampling algorithm provided in an embodiment of the present application;

[0024] Figure 3is a schematic diagram of G-mean experimental results using the Borderline-SMOTE oversampling algorithm provided in an embodiment of the present application;

[0025] Figure 4 is a schematic diagram of G-mean experimental results using the KL-SMOTE oversampling algorithm provided in an embodiment of the present application;

[0026] Figure 5 is a distribution diagram of the original data of random binary classification samples provided in one embodiment of the present application;

[0027] Figure 6 This is a data distribution diagram after SMOTE algorithm sampling provided in an embodiment of the present application;

[0028] Figure 7 It is a data distribution diagram after sampling by the Borderline-SMOTE algorithm provided in an embodiment of the present application;

[0029] Figure 8 This is a data distribution diagram after sampling by the KL-SMOTE algorithm provided in an embodiment of the present application;

[0030] Fig. 9 is a structural diagram of a radar operation data balance classification system based on SMOTE provided in another embodiment of the present application;

[0031] Fig.10 It is a structural schematic diagram of an electronic device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. In the various embodiments of the present application, many technical details are proposed in order to make the reader better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present application can also be implemented. The division of the following embodiments is only for the convenience of description, and the specific implementation mode of the present application should not constitute any limitation. The various embodiments can be combined with each other and referenced to each other under the premise of no contradiction.

[0033] An embodiment of the present application proposes a radar operation data balancing classification method based on SMOTE, which is applied to an electronic device, wherein the electronic device may be a terminal or a server. In this embodiment and the following embodiments, the electronic device is described using a server as an example. The implementation details of the radar operation data balancing classification method based on SMOTE proposed in this embodiment are described in detail below. The following content is only the implementation details provided for ease of understanding and is not necessary for the implementation of this solution.

[0034] The specific process of the radar operation data balance classification method based on SMOTE proposed in this embodiment can be as follows: Figure 1 As shown, including:

[0035] Step 101 , performing classification preprocessing on the acquired radar operation data, and dividing the acquired radar operation data into majority class samples and minority class samples according to the normal operation data and the fault operation data.

[0036] In the specific implementation, the server obtains radar operation data, whose set is denoted as M, and then classifies and preprocesses the obtained radar operation data, dividing it into majority class samples and minority class samples according to normal operation data and fault operation data. The majority class samples are normal operation data, and the minority class samples are fault operation data.

[0037] In one example, the radar operation data includes normal operation data, slight fault operation data, general fault operation data, and severe fault operation data. Generally speaking, the normal operation data includes normal operation data, and the fault operation data includes slight fault operation data, general fault operation data, and severe fault operation data. In some special cases, the normal operation data may include normal operation data and slight fault operation data, and the fault operation data may include general fault operation data and severe fault operation data.

[0038] Step 102, based on the original distribution of the radar operation data, divide each minority class sample into a minority class boundary sample and a minority class internal sample, wherein the minority class boundary sample is a minority class sample distributed at the classification boundary between the majority class and the minority class, and the minority class internal sample is a minority class sample distributed inside the minority class.

[0039] In the specific implementation, after completing the division of the majority class and the minority class, the server needs to further divide the minority class samples into boundaries and interiors. Based on the original distribution of the radar operation data, the server divides the minority class samples into minority class boundary samples and minority class interior samples, where minority class boundary samples are minority class samples distributed at the boundary between the majority class and the minority class, and minority class interior samples are minority class samples distributed within the minority class.

[0040] In one example, after completing the division of the majority class and the minority class, the server records the minority class samples in the radar operation data as c i , i = 1, 2, ..., g, g is the total number of minority class samples. For minority class samples c i , the server calculates c through the Euclidean distance i of Neighbor, record this The number of samples belonging to the majority class among the nearest neighbors And compare and If The server will c i Divided into minority class boundary samples c ib , and store it in the minority class boundary sample set B{c ib}.like The server will c i Divided into minority class internal samples c ii , and stored in the minority class internal sample set I{c ii}.like The server directly sends c i are considered as noise samples and removed.

[0041] Step 103: Process the minority class boundary samples using the KL-SMOTE algorithm to obtain the first class of synthetic samples.

[0042] In a specific implementation, after completing the division of the boundary and interior of the minority class samples, the server can process the minority class boundary samples using the KL-SMOTE algorithm to obtain the first type of synthetic samples.

[0043] In one example, the minority class boundary sample Ci b is stored in the set B{c ib}, for the minority class boundary sample c ib , server computing c ib In the set B{c ib}, and are denoted by c iμ . Then based on c ib and c iμ Interpolate to generate the first type of synthetic samples and will Save to collection

[0044] In an example, the interpolation formula is specifically expressed as:

[0045]

[0046] Among them, random(0,1) means generating a random number between 0 and 1, and the set It represents the set of the first type of synthetic samples.

[0047] At this point, the number of minority class boundary samples has been effectively expanded.

[0048] Step 104: The minority class internal samples are divided into different sub-clusters using the K-means clustering algorithm, and different sampling ratios are calculated according to the number of minority class internal samples in each sub-cluster for synthesis to obtain the second type of synthetic samples.

[0049] In the specific implementation, after completing the boundary and internal division of the minority class samples, the server can use the K-means clustering algorithm to divide the internal samples of the minority class into different sub-clusters, and calculate different sampling multiples according to the number of internal samples of the minority class in each sub-cluster for synthesis to obtain the second type of synthetic samples.

[0050] It is understandable that there is no fixed order between step 103 and step 104. The server may execute step 103 first and then step 104, or execute step 104 first and then step 103, or execute step 103 and step 104 simultaneously.

[0051] In one example, the minority class internal sample c ii Stored in the collection I{c ii}, for the set I{c ii}, the server needs to use the K-means clustering algorithm to divide it into n different sub-clusters (the value of n is generally between 27 and 34), and calculate the cluster center of each sub-cluster. The cluster center of each sub-cluster is recorded as {β 1 , β 2 , ..., β j , ..., β n}, β j is the cluster center of the jth subcluster.

[0052] The server calculates the sampling rate N of each sub-cluster based on the number of samples in the minority class in each sub-cluster. j , sampling rate N j The calculation formula is expressed as:

[0053]

[0054] in, represents the number of samples within the minority class that accounts for the largest proportion in the j-th subcluster, represents the number of samples in the minority class with the smallest proportion in the j-th subcluster, and size(·) represents the size function.

[0055] For the jth sub-cluster, based on the minority class internal samples and β j Synthesize the second type of synthetic samples to obtain the second type of synthetic samples Among them, the number of the second type of synthetic samples that need to be synthesized is N j The product of the number of samples in the minority class in the jth subcluster is expressed as

[0056] The second type of synthetic samples Stored in collection middle, That is, the collection of the second type of synthetic samples.

[0057] At this point, intra-class balance between samples within the minority class is achieved.

[0058] Step 105: merge the first type of synthetic samples, the second type of synthetic samples and the majority type of samples to obtain a balanced sample set.

[0059] In a specific implementation, after obtaining the first type of synthetic samples and the second type of synthetic samples, the server may merge the first type of synthetic samples, the second type of synthetic samples and the majority type samples to obtain a balanced sample set.

[0060] Step 106: Based on the balanced sample set, different classifiers are trained and tested, and radar operation status analysis is performed based on the classification results of each classifier.

[0061] In a specific implementation, after obtaining the balanced sample set, the server can train and test different classifiers based on the balanced sample set, and perform radar operation status analysis based on the classification results of each classifier.

[0062] In one example, different classifiers include GBDT, KNN, and random forest. The server trains and tests different classifiers based on a test set and a balanced sample set, where the test set is obtained by partitioning from radar operation data based on K-fold cross validation.

[0063] In this embodiment, the acquired radar operation data is first divided into majority class and minority class according to the normal operation data and the fault operation data, and then the boundary and internal division of each minority class sample is performed based on the original distribution of the radar operation data. For the minority class boundary samples, the KL-SMOTE algorithm is selected for processing to obtain the first class of synthetic samples to expand the minority class boundary samples. For the minority class internal samples, the K-means clustering algorithm is selected to divide the sub-clusters, and then different sampling multiples are calculated according to the number of minority class internal samples in each sub-cluster for synthesis to obtain the second class of synthetic samples, thereby achieving intra-class balance between minority class internal samples. Finally, the first class of synthetic samples, the second class of synthetic samples and the majority class samples are merged to obtain a balanced sample set to train and test the classifier, and then the radar operation status analysis is performed based on the classification results. After targeted processing, the imbalance problem of minority class samples is effectively avoided, thereby improving the analysis and processing accuracy of radar operation data and reducing the probability of traffic accidents.

[0064] The step division of the above methods is only for clear description. When implemented, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this application; adding insignificant modifications to the algorithm or process or introducing insignificant designs without changing the core design of the algorithm and process are all within the scope of protection of this application.

[0065] In one embodiment, the server imports a radar unbalanced operation data set, including four types of data: ordinary working data, slight fault working data, general fault working data, and severe fault working data. The data set is divided into a training set and a test set based on K-fold cross validation. The training set is divided into a majority class subset and a minority class subset, and the minority class subset is expanded by the KL-SMOTE algorithm, and the other three types of fault operation data except the slight fault working data are expanded to 533 samples, and the other three types of expanded data are combined with the slight fault working data to obtain a balanced data set. Finally, the classifier is trained and tested in the balanced data set and the test set, respectively.

[0066] When selecting the number of clusters for samples within the minority class, the number of clusters, that is, the number of clusters n in the K-means algorithm, is often very important for the sampling effect. Generally speaking, the value of n is selected between 27 and 34. The G-mean and F1-score values ​​of the random forest classifier after data enhancement under different n values ​​are compared. The classification results of KL-SMOTE on the radar operation data set under different n values ​​are shown in Table 1.

[0067] Table 1: Classification results of KL-SMOTE on radar operation dataset with different n values

[0068] The number of clusters n F1-score value G-mean value 27 0.878±0.016 0.836±0.024 28 0.882±0.018 0.841±0.026 29 0.882±0.023 0.841±0.034 30 0.883±0.010 0.843±0.017 31 0.881±0.022 0.839±0.034 32 0.880±0.019 0.838±0.027 33 0.878±0.020 0.836±0.030 34 0.875±0.026 0.829±0.040

[0069] As can be seen from Table 1, when the value of n is 30, the corresponding F1-score is 0.883±0.010 and the G-mean is 0.843±0.017, which are slightly higher than other values ​​of n. Therefore, it can be judged that n=30 is more superior than other values ​​of n. The determination of the number of clusters n depends on the characteristics of the data set itself, that is, the actual distribution of radar operation data, so n=30 is uniformly selected as the optimal number of clusters in the subsequent classifier experimental comparison.

[0070] Next, we conduct comparative experiments on the traditional SMOTE algorithm, the Borderline-SMOTE algorithm, and the KL-SMOTE algorithm in this application. The F1-score values ​​of different SMOTE algorithms under different classifiers are shown in Table 2. The G-mean experimental results using the traditional SMOTE oversampling algorithm can be shown as follows: Figure 2 As shown, the G-mean experimental results using the Borderline-SMOTE oversampling algorithm can be shown as Figure 3 As shown, the G-mean experimental results using the KL-SMOTE oversampling algorithm can be shown as Figure 4 shown.

[0071] Table 2: F1-score values ​​of different SMOTE algorithms under different classifiers

[0072] SMOTE Borderline-SMOTE KL-SMOTE GBDT 0.780 0.787 0.791 KNN 0.733 0.745 0.748 Random Forest 0.817 0.813 0.852

[0073] Based on Table 2, and comparing Figure 2 , Figure 3 , Figure 4 It can be seen that the KL-SMOTE algorithm proposed in this application is significantly better than other algorithms.

[0074] Next, we randomly generate 200 binary classification samples, with the ratio of majority class to minority class being 8:2.

[0075] Figure 5 is the distribution diagram of the original data of random binary classification samples, Figure 6 , Figure 7 , Figure 8 They are the data distribution graph after sampling by the SMOTE algorithm, the data distribution graph after sampling by the Borderline-SMOTE algorithm, and the data distribution graph after sampling by the KL-SMOTE algorithm.

[0076] like Figure 6As shown in the figure, circles represent majority class samples, triangles represent minority class samples, and minority class samples have the characteristics that most samples are concentrated at the classification boundary and the distribution of samples within the class is unbalanced. For the original data set, three oversampling methods are used for comparison and the sampling results are visualized. Figure 7 As shown in the figure, under the traditional SMOTE algorithm, because the intra-class imbalance of minority class samples is not taken into account, random sampling will cause most samples to be distributed around the data with a large proportion in the minority class. Figure 8 As shown in the figure, under the Borderline-SMOTE algorithm, the data near the boundary is sampled. It can be seen that the sampled data is concentrated on the classification boundary, ignoring the contribution of the remaining minority class samples to the classification. Fig. 9 As shown in the figure, under the KL-SMOTE algorithm, the overall distribution of the data after the minority class is sampled is relatively balanced, and there is no obvious sample aggregation area. The sampling near the boundary can also clearly distinguish the majority class from the minority class. Within the minority class, the minority samples far from the boundary are sampled at different sampling rates, so that the number of these minority samples and the number of majority samples are relatively balanced. Therefore, the KL-SMOTE algorithm can improve the recognition rate of sampling points far from the boundary to a certain extent, and at the same time avoid the problem of blurred boundaries.

[0077] Another embodiment of the present application proposes a radar operation data balancing classification system based on SMOTE. The implementation details of the radar operation data balancing classification system based on SMOTE proposed in this embodiment are described in detail below. The following content is only the implementation details provided for the convenience of understanding and is not necessary for the implementation of this example. Fig. 9 It is a structural diagram of a radar operation data balance classification system based on SMOTE proposed in this embodiment, including: a majority-minority division module 201, a boundary internal division module 202, a first type synthesis module 203, a second type synthesis module 2041, a merging module 205 and an analysis and processing module 206.

[0078] The majority-minority classification module 201 is used to perform classification preprocessing on the acquired radar operation data, and to classify the normal operation data and the fault operation data into majority class samples and minority class samples.

[0079] The boundary internal division module 202 is used to divide each minority class sample into minority class boundary samples and minority class internal samples based on the original distribution of the radar operation data, wherein the minority class boundary samples are minority class samples distributed at the classification boundary between the majority class and the minority class, and the minority class internal samples are minority class samples distributed inside the minority class.

[0080] The first type synthesis module 203 is used to process the minority class boundary samples using the KL-SMOTE algorithm to obtain the first type of synthetic samples.

[0081] The second type synthesis module 204 is used to divide the internal samples of the minority class into different sub-clusters using the K-means clustering algorithm, and calculate different sampling multiples according to the number of internal samples of the minority class in each sub-cluster for synthesis to obtain the second type of synthetic samples.

[0082] The merging module 205 is used to merge the first type of synthetic samples, the second type of synthetic samples and the majority type samples to obtain a balanced sample set.

[0083] The analysis and processing module 206 is used to train and test different classifiers based on the balanced sample set, and perform radar operation status analysis based on the classification results of each classifier.

[0084] It is worth mentioning that all modules involved in this embodiment are logic modules. In practical applications, a logic unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application, but this does not mean that there are no other units in this embodiment.

[0085] It is not difficult to find that this embodiment is a system embodiment corresponding to the above method embodiment, and this embodiment can be implemented in conjunction with the above method embodiment. The relevant technical details and technical effects mentioned in the above embodiments are still valid in this embodiment, and in order to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied in the above embodiments.

[0086] Another embodiment of the present application provides an electronic device, whose specific structure is as follows: Fig.10 As shown, it includes: at least one processor 301; and a memory 302 that is communicatively connected to the at least one processor 301; wherein the memory 302 stores instructions that can be executed by the at least one processor 301, and the instructions are executed by the at least one processor 301 so that the at least one processor 301 can execute a radar operation data balancing classification method based on SMOTE described in the above-mentioned method embodiments.

[0087] Among them, the memory and the processor are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be further described in this article. The bus interface is responsible for providing an interface between the bus and the transceiver. The transceiver can be one component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor is transmitted on a wireless medium through an antenna, and further, the antenna also receives data and transmits the data to the processor.

[0088] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0089] Another embodiment of the present application proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement a radar operation data balance classification method based on SMOTE described in the above method embodiments.

[0090] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, ROM (Read-Only Memory), RAM (Random Access Memory), disk or optical disk and other media that can store program codes.

[0091] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A radar operation data balance classification method based on SMOTE, characterized in that: include: The acquired radar operation data is classified and preprocessed, and divided into majority class samples and minority class samples according to normal operation data and fault operation data; Based on the original distribution of radar operation data, each minority class sample is divided into minority class boundary samples and minority class internal samples; among them, minority class boundary samples are minority class samples distributed at the classification boundary between majority class and minority class, and minority class internal samples are minority class samples distributed inside minority class; The KL-SMOTE algorithm is used to process the minority class boundary samples to obtain the first type of synthetic samples; The K-means clustering algorithm is used to divide the internal samples of the minority class into different sub-clusters, and different sampling ratios are calculated according to the number of internal samples of the minority class in each sub-cluster for synthesis, so as to obtain the second type of synthetic samples; The first type of synthetic samples, the second type of synthetic samples and the majority type of samples are combined to obtain a balanced sample set; Based on the balanced sample set, different classifiers are trained and tested, and the radar operation status is analyzed based on the classification results of each classifier; The KL-SMOTE algorithm is used to process the minority class boundary samples to obtain the first type of synthetic samples, including: For minority class boundary samples ,calculate In the collection In neighbors, and are denoted as ; based on and Interpolate to generate the first type of synthetic samples , and Save to collection The interpolation formula is specifically expressed as , Indicates the generation of a random number between 0 and 1, the set That is, the set of the first type of synthetic samples; The K-means clustering algorithm is used to divide the internal samples of the minority class into different sub-clusters, and different sampling ratios are calculated according to the number of internal samples of the minority class in each sub-cluster for synthesis, and the second type of synthetic samples are obtained, including: For collections , using K-means clustering algorithm to divide into different sub-clusters and calculate the cluster center of each sub-cluster, denoted as , For the The cluster center of each sub-cluster; The sampling ratio of each sub-cluster is calculated based on the number of samples within the minority class in each sub-cluster. , sampling rate The calculation formula is expressed as: ； in, Indicates The number of samples within the minority class that accounts for the largest proportion in each sub-cluster, Indicates The number of samples within the minority class that accounts for the smallest proportion in each sub-cluster, represents the size function; For For each sub-cluster, based on the minority class internal samples and Perform synthesis of the second type of synthetic samples to obtain the second type of synthetic samples ; Among them, the number of the second type of synthetic samples that need to be synthesized is With The product of the number of samples in the minority class in the sub-clusters, the synthesis formula is specifically expressed as .

2. The radar operation data balance classification method based on SMOTE as claimed in claim 1, characterized in that: The minority class samples in the radar operation data are denoted as , , is the total number of minority class samples. Based on the original distribution of radar operation data, each minority class sample is divided into minority class boundary samples and minority class internal samples, including: For minority class samples , calculated by Euclidean distance of Neighbor, record this The number of samples belonging to the majority class among the nearest neighbors , and compare and size; like , then Classify as minority class boundary samples , and store it in a collection ; like , then Classify as minority class internal samples , and store it in a collection ; like , then are considered as noise samples and removed.

3. The radar operation data balance classification method based on SMOTE as claimed in claim 1, characterized in that: The value range is 27 to 34.

4. The radar operation data balancing classification method based on SMOTE according to any one of claims 1 to 3, characterized in that: The radar operation data includes normal operation data, minor fault operation data, general fault operation data, and serious fault operation data; the normal operation data includes normal operation data; the fault operation data includes minor fault operation data, general fault operation data, and serious fault operation data; Alternatively, the normal operation data includes ordinary operation data and minor fault operation data, and the fault operation data includes general fault operation data and major fault operation data.

5. The radar operation data balancing classification method based on SMOTE according to any one of claims 1 to 3, characterized in that: Different classifiers include GBDT, KNN, and random forest; Based on the balanced sample set, different classifiers are trained and tested, including: Different classifiers are trained and tested based on the test set and the balanced sample set. The test set is obtained by dividing the radar operation data based on K-fold cross-validation.

6. A radar operation data balance classification system based on SMOTE, characterized in that: include: The majority-minority classification module is used to classify and pre-process the acquired radar operation data, and divide it into majority class samples and minority class samples according to normal operation data and fault operation data; The boundary internal division module is used to divide each minority class sample into minority class boundary samples and minority class internal samples based on the original distribution of radar operation data, wherein the minority class boundary samples are minority class samples distributed at the classification boundary between the majority class and the minority class, and the minority class internal samples are minority class samples distributed within the minority class; The first type synthesis module is used to process the minority class boundary samples using the KL-SMOTE algorithm to obtain the first type of synthetic samples; The second type of synthesis module is used to divide the internal samples of the minority class into different sub-clusters using the K-means clustering algorithm, and calculate different sampling multiples according to the number of internal samples of the minority class in each sub-cluster for synthesis to obtain the second type of synthetic samples; A merging module, used for merging the first type of synthetic samples, the second type of synthetic samples and the majority type of samples to obtain a balanced sample set; An analysis and processing module is used to train and test different classifiers based on a balanced sample set, and to analyze the radar operation status based on the classification results of each classifier; The KL-SMOTE algorithm is used to process the minority class boundary samples to obtain the first type of synthetic samples, including: For minority class boundary samples ,calculate In the collection In neighbors, and are denoted as ; based on and Interpolate to generate the first type of synthetic samples , and Save to collection The interpolation formula is specifically expressed as , Indicates the generation of a random number between 0 and 1, the set That is, the set of the first type of synthetic samples; The K-means clustering algorithm is used to divide the internal samples of the minority class into different sub-clusters, and different sampling ratios are calculated according to the number of internal samples of the minority class in each sub-cluster for synthesis, and the second type of synthetic samples are obtained, including: For collections , using K-means clustering algorithm to divide into different sub-clusters and calculate the cluster center of each sub-cluster, denoted as , For the The cluster center of each sub-cluster; The sampling ratio of each sub-cluster is calculated based on the number of samples within the minority class in each sub-cluster. , sampling rate The calculation formula is expressed as: ； in, Indicates The number of samples within the minority class that accounts for the largest proportion in each sub-cluster, Indicates The number of samples within the minority class that accounts for the smallest proportion in each sub-cluster, represents the size function; For For each sub-cluster, based on the minority class internal samples and Perform synthesis of the second type of synthetic samples to obtain the second type of synthetic samples ; Among them, the number of the second type of synthetic samples that need to be synthesized is With The product of the number of samples in the minority class in the sub-clusters, the synthesis formula is specifically expressed as .

7. An electronic device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; Wherein, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the SMOTE-based radar operation data balancing classification method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the radar operation data balancing classification method based on SMOTE as described in any one of claims 1 to 5 can be implemented.

Images