Target feature clustering method and device
By calculating the correlation between the target category and the target feature in the K-Means algorithm, generating target features with increased distances, and determining clustering categories based on the correlation and distance of these features, the problem of inaccurate classification in dense data is solved, and more accurate clustering classification is achieved.
Patent Information
- Application Number
- CN202411763885.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-03
- Publication Date
- 2025-04-22
AI Technical Summary
The K-Means algorithm is prone to inaccurate classification when processing dense data, especially for data at the cluster edge, which is difficult to accurately assign it to appropriate clusters.
By calculating the correlation between the target category and the first target feature, weights are configured for the first target feature, and the second target feature is generated, so that the distance between it and other features is increased; then, the correlation between the plurality of second target features is calculated, the order is sorted, and the weight is configured for the second target feature based on the sorting result, and the third target feature is generated; finally, the target category to which the first target feature belongs is determined based on the distance between the plurality of third target features and the target category.
Improve the accuracy of clustering classification, especially when processing dense data, effectively avoiding the problem of inaccurate classification and achieving a more accurate classification of target data.
Smart Images
Figure CN119862438A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data mining, and in particular to a target feature clustering method and device. Background Art
[0002] Clustering classification technology is an important technology in the field of data mining. Its main purpose is to divide the data points in the data set into several categories according to their similarities. With the development of data mining technology, clustering classification technology has been widely used in various fields, such as market analysis, social network analysis, image segmentation, etc. The development of clustering classification technology has gone through a process from simple to complex, from static to dynamic, and from single to multiple. Early clustering classification technology was mainly based on distance measurement, such as the K-Means algorithm. With the development of technology, clustering classification technology has gradually introduced probability models, neural networks, deep learning and other methods, making clustering classification technology more accurate and efficient.
[0003] In related technologies, the K-Means algorithm is a commonly used clustering classification method. Its basic idea is to assign each data point in a data set to the nearest cluster center, and then update the location of the cluster center until a certain stop condition is met. The K-Means algorithm is simple and efficient, and is suitable for clustering classification of large-scale data sets.
[0004] However, in practical applications, the K-Means algorithm has certain limitations. When there are some relatively dense data in the data set, the K-Means algorithm is prone to inaccurate classification. Especially for some relatively concentrated data at the edge of the cluster, the distance between the data is very close, and it is difficult to accurately assign these data to the appropriate cluster. Summary of the invention
[0005] The embodiments of the present invention provide a target feature clustering method and device, which at least solve the problem in the related art that some concentrated data is difficult to be divided into appropriate clusters.
[0006] According to an embodiment of the present invention, a target feature clustering method is provided, comprising:
[0007] Calculating the correlation between the target category and the first target feature;
[0008] assigning weights to the first target features based on the association to obtain second target features, so that distances between a plurality of the second target features increase;
[0009] Calculating the correlation between the plurality of the second target features to sort the plurality of the second target features;
[0010] configuring a weight for the second target feature based on the sorting result to obtain a third target feature;
[0011] The target category to which the first target feature belongs is determined based on the distances between the plurality of third target features and the target category.
[0012] Optionally, before calculating the correlation between the plurality of second target features to sort the plurality of second target features, the method further includes:
[0013] Determine a first set of the second target features;
[0014] The second target features in the first set are checked item by item based on a preset optimization algorithm to delete outliers.
[0015] Optionally, before determining the target category to which the first target feature belongs based on the distances between the plurality of third target features and the target category, the method further includes:
[0016] Determine a second set obtained by the third target feature;
[0017] The third target feature in the second set is checked based on a preset optimization algorithm to delete outliers.
[0018] Optionally, the preset optimization algorithm includes:
[0019] Calculate a first quantity of all the second target features in the first set;
[0020] Calculating a second number of the second target features that are identical in the first set;
[0021] When the ratio of the second number to the first number is greater than a first threshold, deleting the second target feature corresponding to the second number;
[0022] or,
[0023] Calculating a third quantity of all the third target features in the second set;
[0024] Calculating a fourth number of the third target features that are identical in the second set;
[0025] When the ratio of the fourth number to the third number is greater than a first threshold, the third target feature corresponding to the fourth number is deleted.
[0026] Optionally, the preset optimization algorithm includes:
[0027] Outliers are eliminated based on the data box plot of the second target feature or the third target feature.
[0028] Optionally, the preset optimization algorithm includes:
[0029] Calculating the square difference between the second target features, and deleting the second target feature when the square difference is less than or equal to a preset second threshold;
[0030] or,
[0031] The square difference between the third target features is calculated, and when the square difference is less than or equal to a preset second threshold, the third target feature is deleted.
[0032] Optionally, after determining the target category to which the first target feature belongs based on the distances between the plurality of third target features and the target category, the method further includes:
[0033] In the case that the target category includes a plurality of the first target features, a vector mean of the plurality of third target features is calculated, and the vector mean is used as a new target category, wherein the third target feature corresponds to the first target feature.
[0034] According to another embodiment of the present invention, a target feature clustering system is provided, including:
[0035] A first weight configuration module is used to determine the correlation between the target category and the first target feature; and
[0036] assigning weights to the first target features based on the association to obtain second target features, so that distances between a plurality of the second target features increase;
[0037] a second weight configuration module, configured to calculate the correlation between the plurality of the second target features so as to sort the plurality of the second target features; and
[0038] configuring a weight for the second target feature based on the sorting result to obtain a third target feature;
[0039] An execution module is used to determine the target category to which the first target feature belongs based on the distances between the multiple third target features and the target category.
[0040] According to yet another embodiment of the present invention, a computer-readable storage medium is provided, wherein a computer program is stored in the storage medium, wherein the computer program executes the steps in any of the above method embodiments when executed by a processor.
[0041] According to yet another embodiment of the present invention, there is provided an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.
[0042] According to one of the embodiments of the present invention, first, by calculating the correlation between the target category and the first target feature, and configuring the weight for the first target feature based on these correlations, the second target feature is generated. This method can increase the distance between multiple second target features, so that when facing dense data, the distance between dense data is appropriately increased, for example, "dense data" is adjusted to "sparse data", and cluster classification is more accurate. Secondly, by calculating the correlation between multiple second target features, they are sorted, and based on the sorting result, the weight is configured for the second target feature to obtain the third target feature. This method optimizes the cluster classification process, making the selection of cluster centers more accurate. Finally, the target category to which the first target feature belongs is determined based on the distance between multiple third target features and the target category. This method not only considers the correlation between features and categories, but also considers the relative distance between features, so that cluster classification can be performed more accurately when processing concentrated data. Therefore, the present invention improves the accuracy of cluster classification through these technical means, especially when processing dense data, it can effectively avoid the problem of inaccurate classification and achieve more accurate classification of target data. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:
[0044] Figure 1 is a flow chart of a target feature clustering method according to an embodiment of the present invention;
[0045] Figure 2 is a schematic diagram of a change in a second target feature obtained by configuring a weight for the first target feature according to an embodiment of the present invention, or a schematic diagram of a change in a third target feature obtained by configuring a weight for the second target feature;
[0046] Figure 3 is a schematic diagram of changes in target features in the first target category and the second target category after weights are configured according to an embodiment of the present invention;
[0047] Figure 4 is a flowchart of a method for deleting outliers in a second target feature according to an embodiment of the present invention;
[0048] Figure 5is a flowchart of a method for deleting outliers in a third target feature according to an embodiment of the present invention;
[0049] Figure 6 is a schematic diagram of changes in SSE during multiple clustering processes according to an embodiment of the present invention;
[0050] Figure 7 It is a structural diagram of a target feature clustering system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0051] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.
[0052] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.
[0053] In this embodiment, a target feature clustering method is provided. Figure 1 is a flow chart of a target feature clustering method according to an embodiment of the present invention. Figure 1 As shown, the process includes:
[0054] Step S101, calculating the correlation between the target category and the first target feature;
[0055] In an exemplary embodiment, for example, the correlation between the first target feature and the target category is determined based on feature matching, similarity calculation, classification algorithm, manual review, etc. The classification algorithm may use an unsupervised learning algorithm, such as K-means clustering, to discover natural groups in the data.
[0056] Step S102, configuring a weight for the first target feature based on the correlation to obtain a second target feature, so that the distance between a plurality of the second target features increases;
[0057] In an exemplary embodiment, for example, in a clustering task, there are 7 target categories and 700 first target features, and the correlations between the 700 first target features and the 7 target categories are calculated respectively. For example, the correlation between the first target feature A and the target category A is 99%, the correlation between the first target feature B and the target category B is 90%, the correlation between the first target feature C and the target category C is 88%, ..., the correlation between the first target feature F and the target category F is 25%, for example, the correlation of 99% is the maximum value among the seven groups of data, then the first weight is configured for the first target feature A, so that the first target feature A is "close" to the target category A in the space vector.
[0058] Step S103, calculating the correlation between the plurality of second target features to sort the plurality of second target features;
[0059] In an exemplary implementation, for example, the similarity between multiple second target features may be calculated. By calculating the similarity between the second target features, it may be determined whether the multiple second target features belong to the same target category.
[0060] Step S104, configuring a weight for the second target feature based on the sorting result to obtain a third target feature;
[0061] For example, in a clustering task, the correlation between the first target feature A and the first target feature B is 98%, the correlation between the first target feature A and the first target feature C is 50%, the correlation between the first target feature A and the first target feature D is 20%, ..., the correlation between the first target feature A and the first target feature F is 5%. For example, if the correlation is 98%, which is the maximum value among the seven groups of data, it is determined that the correlation between the first target feature A and the first target feature B is the largest, and the second weight is configured for the first target feature A so that the first target feature A is "close to" the first target feature B in the space vector. Or if the correlation is 5%, which is the minimum value among the seven groups of data, it is determined that the correlation between the first target feature A and the first target feature F is the smallest, and the third weight is configured for the first target feature A so that the first target feature A is "far away from" the first target feature F in the space vector.
[0062] In an exemplary embodiment, Figure 2 is a schematic diagram of a change in a second target feature obtained by configuring a weight for the first target feature according to an embodiment of the present invention, or a schematic diagram of a change in a third target feature obtained by configuring a weight for the second target feature, such as Figure 2 As shown, for example, in the five groups of first target features in a clustering process, the five groups of first target features are vectorized into: vector A, vector B, vector C, vector D, vector E (for example, the left figure), and respective weights are assigned to vector A, vector B, vector C, vector D, and vector E, respectively, to obtain vector A1, vector B1, vector C1, vector D1, and vector E1 (for example, the right figure), that is, to obtain the second target feature, so that the distance between the corresponding multiple first target features is increased. Alternatively, vector A, vector B, vector C, vector D, and vector E can also correspond to the second target feature, and the respective weights are assigned to the second target feature, and the third target feature has been obtained: vector A1, vector B1, vector C1, vector D1, and vector E1, so that the distance between the corresponding multiple first target features is increased.
[0063] In an exemplary embodiment, Figure 3is a schematic diagram of changes in target features in the first target category and the second target category after being configured with weights according to an embodiment of the present invention, such as Figure 3 As shown, there are some edge data in the edge area adjacent to the first target category set and the second target category set in the left figure, such as vector A, vector B, vector C, vector D, and vector E. After the weights are configured in step S102 and / or step S104, the situation changes to the right figure. After the weight configuration process, the edge data in the first target category set is closer to the first target category, and the edge data in the second target category set is closer to the second target category, so that the distance between the edge data in the first target category set and the edge data in the second target category set is "increased", so that the difference between the edge data in the first target category set and the edge data in the second target category set is increased, so that the cluster classification result is more accurate.
[0064] Step S105: determining the target category to which the first target feature belongs based on the distances between the plurality of third target features and the target category.
[0065] In an exemplary implementation, for example, the Euclidean distance between the third target feature and the target category may be calculated to determine the relationship between the third target feature and the target category.
[0066] Through step S101 to step S105, first, by calculating the correlation between the target category and the first target feature, and configuring weights for the first target feature based on these correlations, the second target feature is generated. This method can increase the distance between multiple second target features, so that when facing dense data, the distance between dense data is appropriately increased, for example, adjusting "dense data" to "sparse data", clustering classification is more accurate.
[0067] Secondly, by calculating the correlation between multiple second target features, sorting them, and configuring weights for the second target features based on the sorting results, the third target feature is obtained. This method optimizes the clustering classification process and makes the selection of cluster centers more accurate.
[0068] Finally, the target category to which the first target feature belongs is determined based on the distances between the multiple third target features and the target categories. This method not only considers the correlation between features and categories, but also the relative distances between features, so that clustering and classification can be performed more accurately when processing centralized data.
[0069] Therefore, the present invention improves the accuracy of cluster classification through these technical means, especially when processing dense data, it can effectively avoid the problem of inaccurate classification and achieve more accurate classification of target data.
[0070] In one embodiment, Figure 4 is a flowchart of a method for deleting outliers in a second target feature according to an embodiment of the present invention. Figure 4 As shown, before calculating the correlation between the plurality of second target features to sort the plurality of second target features, the method further includes:
[0071] Step S401, determining a first set obtained by the second target feature;
[0072] Step S402: Check the second target features in the first set one by one based on a preset optimization algorithm to delete abnormal values.
[0073] In one embodiment, Figure 5 is a flowchart of a method for deleting outliers in a third target feature according to an embodiment of the present invention. Figure 5 As shown, before determining the target category to which the first target feature belongs based on the distances between the plurality of third target features and the target category, the method further includes:
[0074] Step S501, determining a second set obtained by the third target feature;
[0075] Step S502: Check the third target feature in the second set based on a preset optimization algorithm to delete outliers.
[0076] In one embodiment, the preset optimization algorithm includes:
[0077] Calculate a first quantity of all the second target features in the first set;
[0078] Calculating a second number of the second target features that are identical in the first set;
[0079] When the ratio of the second number to the first number is greater than a first threshold, deleting the second target feature corresponding to the second number;
[0080] or,
[0081] Calculating a third quantity of all the third target features in the second set;
[0082] Calculating a fourth number of the third target features that are identical in the second set;
[0083] When the ratio of the fourth number to the third number is greater than a first threshold, the third target feature corresponding to the fourth number is deleted.
[0084] In an exemplary embodiment, for example, the first threshold value can be 70%. The first number of all second target features in the first set is 100, and the second number of identical second target features in the first set is 85. The ratio of the second number to the first number is 85%, which is greater than 70%. In this case, the 85 identical second target features can be deleted from the set to obtain a new first set. Similarly, the third number of all third target features in the second set is 100, and the fourth number of identical third target features in the second set is 75. The ratio of the fourth number to the third number is 75%, which is greater than 70%. In this case, the 75 identical third target features can be deleted from the set to obtain a new second set.
[0085] In one embodiment, the preset optimization algorithm includes:
[0086] Outliers are eliminated based on the data box plot of the second target feature or the third target feature.
[0087] In an exemplary embodiment, a box plot, also known as a box-and-whisker plot, a box plot or a box-and-line plot, is a statistical chart used to display data distribution characteristics. It describes the distribution of data through five key values (minimum value, first quartile, median, third quartile and maximum value), which are also called the "five-number summary" of the box plot.
[0088] The basic structure of a box plot is as follows:
[0089] Box: represents the middle 50% of the data, that is, the range between the first quartile (Q1) and the third quartile (Q3).
[0090] Box lines: The upper and lower boundaries of the box, representing Q1 and Q3 respectively.
[0091] Center Line: The line inside the box represents the median of the data (Q2).
[0092] Whiskers: Lines extending from the box represent the minimum and maximum values of the data, but the length of the whiskers usually does not exceed 1.5 times the IQR (interquartile range, i.e. Q3-Q1).
[0093] Outliers: Data points that fall outside the whiskers, usually represented by dots or asterisks.
[0094] Methods for identifying outliers from box plots:
[0095] Calculate IQR: IQR = Q3 - Q1.
[0096] Determine the limits of outliers: lower limit = Q1-1.5*IQR, upper limit = Q3+1.5*IQR.
[0097] Observe Data Points: Data points outside the lower and upper bounds are considered outliers.
[0098] Through the box plot, you can intuitively understand the distribution characteristics of the data, including central tendency, dispersion, and outliers.
[0099] In one embodiment, the preset optimization algorithm includes:
[0100] Calculating the square difference between the second target features, and deleting the second target feature when the square difference is less than or equal to a preset second threshold;
[0101] or,
[0102] The square difference between the third target features is calculated, and when the square difference is less than or equal to a preset second threshold, the third target feature is deleted.
[0103] In an exemplary embodiment, the square difference between multiple second target features is calculated, and the similarity between the two can be determined. For example, the second threshold is 2. When the square difference between the second target feature A and the second target feature B is 1, which is less than the second threshold, it is determined that the second target feature A and the second target feature B are the same value, and the second target feature A and the second target feature B are deleted. Similarly, when the square difference between the third target feature A and the third target feature B is 1, which is less than the second threshold, it is determined that the third target feature A and the third target feature B are the same value, and the third target feature A and the third target feature B are deleted.
[0104] In one implementation, after determining the target category to which the first target feature belongs based on the distances between the plurality of third target features and the target category, the method further includes:
[0105] In the case that the target category includes a plurality of the first target features, a vector mean of the plurality of third target features is calculated, and the vector mean is used as a new target category, wherein the third target feature corresponds to the first target feature.
[0106] In an exemplary implementation, the target feature may be vectorized, and the calculation formula of the vectorized vector mean V(X, Ci) is as follows:
[0107]
[0108] Among them, X is the data object, that is, the target feature. Ci is the i-th cluster center, that is, the i-th target category; m is the dimension of the data object; Xj, Ci j are the j-th attribute values of X and Ci respectively.
[0109] In an exemplary implementation, the Euclidean distance between the third target feature and the target category may be calculated to determine the relationship between the third target feature and the target category.
[0110] Among them, the calculation formula of Euclidean distance can be:
[0111]
[0112] In an exemplary implementation, in order to verify whether the above clustering results meet the requirements, the error sum of squares SSE of the entire data set can be calculated. The SSE can be calculated using the following formula:
[0113]
[0114] In one embodiment, Figure 6 is a schematic diagram of the change of SSE in multiple clustering processes according to an embodiment of the present invention, such as Figure 6 As shown in the figure, for example, in a clustering task, N clustering processes are performed. Among them, the SSE value changes the most before and after clustering the third feature target, and it can be considered that the clustering result of this process meets the requirements.
[0115] It should be noted that the above method can also be applied to the application scenario of screening target customers for product managers, which is of course only an exemplary application and the present invention is not limited thereto. For example, the above target feature can be a target customer, and the above target category can be a product planned to be recommended to the customer.
[0116] In this embodiment, a target feature clustering system is also provided. Figure 7 is a structural diagram of the target feature clustering system of an embodiment of the present invention, such as Figure 7 As shown, the target feature clustering system includes:
[0117] A first weight configuration module 71 is used to determine the correlation between the target category and the first target feature; and
[0118] assigning weights to the first target features based on the association to obtain second target features, so that distances between a plurality of the second target features increase;
[0119] A second weight configuration module 72 is used to calculate the correlation between the plurality of the second target features to sort the plurality of the second target features; and
[0120] configuring a weight for the second target feature based on the sorting result to obtain a third target feature;
[0121] The execution module 73 is configured to determine the target category to which the first target feature belongs based on the distances between the plurality of third target features and the target category.
[0122] By adopting the above technical solution, first, the first weight configuration module 71 is used to calculate the correlation between the target category and the first target feature, and the weight is configured for the first target feature based on these correlations to generate the second target feature. This method can increase the distance between multiple second target features, so that when facing dense data, the distance between dense data is appropriately increased, for example, "dense data" is adjusted to "sparse data", and cluster classification is more accurate.
[0123] Secondly, the second weight configuration module 72 is used to calculate the correlation between the plurality of second target features, sort them, and configure weights for the second target features based on the sorting results to obtain the third target feature. This method optimizes the clustering classification process and makes the selection of cluster centers more accurate.
[0124] Finally, the target category to which the first target feature belongs is determined based on the distances between the plurality of third target features and the target category by using the execution module 73. This method not only considers the correlation between the feature and the category, but also considers the relative distance between the features, so that clustering and classification can be performed more accurately when processing centralized data.
[0125] Therefore, the present invention improves the accuracy of cluster classification through these technical means, especially when processing dense data, it can effectively avoid the problem of inaccurate classification and achieve more accurate classification of target data.
[0126] In one embodiment, the system is also used to:
[0127] Determine a first set of the second target features;
[0128] The second target features in the first set are checked item by item based on a preset optimization algorithm to delete outliers.
[0129] In one embodiment, the system is also used to:
[0130] Determine a second set obtained by the third target feature;
[0131] The third target feature in the second set is checked based on a preset optimization algorithm to delete outliers.
[0132] In one embodiment, the system is also used to:
[0133] Calculate a first quantity of all the second target features in the first set;
[0134] Calculating a second number of the second target features that are identical in the first set;
[0135] When the ratio of the second number to the first number is greater than a first threshold, deleting the second target feature corresponding to the second number;
[0136] or,
[0137] Calculating a third quantity of all the third target features in the second set;
[0138] Calculating a fourth number of the third target features that are identical in the second set;
[0139] When the ratio of the fourth number to the third number is greater than a first threshold, the third target feature corresponding to the fourth number is deleted.
[0140] In one embodiment, the system is also used to:
[0141] Outliers are eliminated based on the data box plot of the second target feature or the third target feature.
[0142] In one embodiment, the system is also used to:
[0143] Calculating the square difference between the second target features, and deleting the second target feature when the square difference is less than or equal to a preset second threshold;
[0144] or,
[0145] The square difference between the third target features is calculated, and when the square difference is less than or equal to a preset second threshold, the third target feature is deleted.
[0146] In one embodiment, the system is also used to:
[0147] In the case that the target category includes a plurality of the first target features, a vector mean of the plurality of third target features is calculated, and the vector mean is used as a new target category, wherein the third target feature corresponds to the first target feature.
[0148] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.
[0149] An embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above method embodiments are executed.
[0150] In one implementation, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store computer programs.
[0151] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.
[0152] In one implementation manner, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation manners, and this embodiment will not be described in detail herein.
[0153] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.
[0154] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A target feature clustering method, characterized in that: include: Calculating the correlation between the target category and the first target feature; assigning weights to the first target features based on the association to obtain second target features, so that distances between a plurality of the second target features increase; Calculating the correlation between the plurality of the second target features to sort the plurality of the second target features; configuring a weight for the second target feature based on the sorting result to obtain a third target feature; The target category to which the first target feature belongs is determined based on the distances between the plurality of third target features and the target category.
2. The method according to claim 1, characterized in that Before calculating the correlation between the plurality of second target features to sort the plurality of second target features, the method further includes: Determine a first set of the second target features; The second target features in the first set are checked item by item based on a preset optimization algorithm to delete outliers.
3. The method according to claim 1, characterized in that Before determining the target category to which the first target feature belongs based on the distances between the plurality of third target features and the target category, the method further includes: Determine a second set obtained by the third target feature; The third target feature in the second set is checked based on a preset optimization algorithm to delete outliers.
4. The method according to claim 2 or 3, characterized in that: The preset optimization algorithm includes: Calculate a first quantity of all the second target features in the first set; Calculating a second number of the second target features that are identical in the first set; When the ratio of the second number to the first number is greater than a first threshold, deleting the second target feature corresponding to the second number; or, Calculating a third quantity of all the third target features in the second set; Calculating a fourth number of the third target features that are identical in the second set; When the ratio of the fourth number to the third number is greater than a first threshold, the third target feature corresponding to the fourth number is deleted.
5. The method according to claim 2 or 3, characterized in that: The preset optimization algorithm includes: Outliers are eliminated based on the data box plot of the second target feature or the third target feature.
6. The method according to claim 2 or 3, characterized in that: The preset optimization algorithm includes: Calculating the square difference between the second target features, and deleting the second target feature when the square difference is less than or equal to a preset second threshold; or, The square difference between the third target features is calculated, and when the square difference is less than or equal to a preset second threshold, the third target feature is deleted.
7. The method according to claim 1, characterized in that After determining the target category to which the first target feature belongs based on the distances between the plurality of third target features and the target category, the method further includes: In the case that the target category includes a plurality of the first target features, a vector mean of the plurality of third target features is calculated, and the vector mean is used as a new target category, wherein the third target feature corresponds to the first target feature.
8. A target feature clustering system, characterized in that: include: A first weight configuration module, used to determine the correlation between the target category and the first target feature; as well as, assigning weights to the first target features based on the association to obtain second target features, so that distances between a plurality of the second target features increase; A second weight configuration module, used for calculating the correlation between the plurality of the second target features, so as to sort the plurality of the second target features; as well as, configuring a weight for the second target feature based on the sorting result to obtain a third target feature; An execution module is used to determine the target category to which the first target feature belongs based on the distances between the multiple third target features and the target category.
9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program executes the method described in any one of claims 1 to 7 when executed by a processor.
10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 7.