High-dimensional data clustering method and device based on random mapping, terminal equipment and storage medium

By performing dimensionality reduction and random mapping processing on high-dimensional data, the problem of insufficient privacy data protection in the clustering of high-dimensional data in the existing technology is solved, and more accurate and secure clustering results are achieved.

CN118965032AActive Publication Date: 2024-11-15GUANGDONG POWER GRID CO LTD +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411097607.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-11-15
Estimated Expiration
2044-08-12

AI Technical Summary

Technical Problem

The existing privacy protection algorithm cannot effectively protect private data when clustering high-dimensional data, and is easily affected by redundant features, resulting in local optimal problems in clustering results, and the clustering center is exposed, which in turn endangers sensitive information in the original data. .

Method used

By reducing the dimensionality of high-dimensional data, redundant features are removed, and a random mapping matrix is ​​generated using independent and homogeneous Gaussian random variables, random mapping transformation of the dimensionality reduction data is performed to generate data to be clustered, avoid interference from redundant features, and enhance the Privacy data protection.

Benefits of technology

It effectively reduces the impact of redundant features on clustering, avoids local optimal problems, improves the protection effect of private data, and ensures the security of clustering centers.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965032A_ABST
    Figure CN118965032A_ABST
Patent Text Reader

Abstract

The invention discloses a high-dimensional data clustering method and device based on random mapping, terminal equipment and a storage medium, according to the method, data dimension reduction operation is performed on high-dimensional data according to a preset dependent variable, only features related to the dependent variable in the high-dimensional data are reserved, redundant features irrelevant to the dependent variable are removed, and the high-dimensional data clustering efficiency is improved. And the influence of redundant features on subsequent data clustering is greatly reduced. And then, before clustering, carrying out random transformation on the dimension-reduced data by adopting a random mapping matrix so as to achieve the effect of protecting the dimension-reduced data. Therefore, according to the method, the interference of redundant features and the problem of clustering local optimum are avoided by removing the redundant features, and the privacy data are further protected through the random mapping matrix.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of privacy protection data mining technology, and in particular to a high-dimensional data clustering method, device, terminal equipment and storage medium based on random mapping. Background Art

[0002] In recent years, with the rise of big data technology and the rapid development of information technology, various organizations, such as medical institutions, educational institutions, etc., generate a large amount of high-dimensional data every day. Using data mining technology, valuable data can be extracted from this high-dimensional data and used to make scientific decisions. Clustering, as the most effective algorithm used in data mining, has been widely used. However, with the gradual maturity of cluster mining technology in the field of high-dimensional data, cluster analysis may involve a large amount of sensitive information, such as personal information, trade secrets, etc. Therefore, with the development of computing technology and data mining, in information security applications, not only is it necessary to have a generator that can output various forms of random numbers, but also to conduct more in-depth research and analysis on privacy-preserving clustering to provide theoretical support for related applications.

[0003] At present, with the rapid development of information technology, people's security needs for high-security clustering data are gradually increasing. When clustering high-dimensional data, the current privacy protection algorithm still cannot effectively protect the privacy data in high-dimensional data. Since there are many features in high-dimensional data, the current privacy protection algorithm is easily affected by redundant features, which leads to the problem of local optimality in its clustering results, and ultimately makes the generated cluster centers easy to be exposed. When the cluster centers are exposed, attackers can use the information of these cluster centers to obtain sensitive information in the original data.

[0004] Therefore, how to effectively protect the privacy of high-dimensional data has become an urgent problem that needs to be solved. Summary of the invention

[0005] The embodiments of the present invention provide a high-dimensional data clustering method, apparatus, terminal device and storage medium based on random mapping. By reducing the dimensionality of high-dimensional data before clustering and using a random mapping matrix to protect privacy data, the privacy data contained in the high-dimensional data can be protected when the high-dimensional data is clustered.

[0006] An embodiment of the present invention provides a high-dimensional data clustering method based on random mapping, comprising:

[0007] Obtain a high-dimensional data set consisting of several features and several preset dependent variables;

[0008] Determining a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generating reduced-dimensionality data according to the number of relevant features;

[0009] generating a first random mapping matrix according to the number of relevant features in the dimensionally reduced data;

[0010] Wherein, the elements in the first random mapping matrix are all independent and identically distributed Gaussian random variables;

[0011] According to the first random mapping matrix, performing random mapping transformation on the dimension reduction data to generate a feature transformation matrix;

[0012] Generate a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix;

[0013] Using the second random mapping matrix to perform random mapping transformation on the dimension reduction data to generate data to be clustered;

[0014] The data to be clustered is taken as an initial cluster center, and the clustering operation is then repeated. When the number of cluster centers generated by the clustering operation is not less than a preset cluster center target value, a clustering result of the high-dimensional data set is generated.

[0015] Further, determining a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generating dimensionality reduction data according to the number of relevant features, includes:

[0016] Using a symmetric uncertainty formula, calculate the correlation score between each dependent variable and each feature in the high-dimensional data set;

[0017] Determining a number of initial redundant features and a number of initial relevant features according to the relevance score;

[0018] The Pearson correlation coefficient formula is used to calculate the Pearson coefficient between each dependent variable and each initial correlation feature;

[0019] The initial relevant features that are less than a preset coefficient threshold are used as redundant features, and the initial relevant features that are not less than the preset coefficient threshold are used as relevant features, and the dimension reduction data are generated according to the relevant features.

[0020] Furthermore, after determining a plurality of initial redundant features and a plurality of initial relevant features, the method further includes:

[0021] Normalizing the initial data under each initial relevant feature to generate normalized data under each initial relevant feature;

[0022] According to the normalized data under each initial correlation feature, the normal distribution function of each initial correlation feature is plotted, and the standard deviation of each state distribution function is calculated;

[0023] Get the preset data filtering rules;

[0024] When it is determined that the data screening rule is a moderate rule, the normalized data that is greater than two times the standard deviation is deleted;

[0025] When the data screening rule is determined to be a strict rule, the normalized data that is larger than the standard deviation is deleted.

[0026] Furthermore, the feature transformation matrix includes: a feature matching matrix and a noise interference matrix to be added;

[0027] The step of performing random mapping transformation on the dimension reduction data according to the first random mapping matrix to generate a feature transformation matrix includes:

[0028] Orthogonalizing the first random mapping matrix, and then splitting the orthogonalized first random mapping matrix into a first random mapping sub-matrix and a second random mapping sub-matrix;

[0029] The first random mapping sub-matrix and the second random mapping sub-matrix are used to perform random mapping transformation on the dimension reduction data to generate a feature matching matrix and an interference matrix to be noisy.

[0030] Furthermore, generating a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix includes:

[0031] Performing a noise adding operation on the interference matrix to be noised to generate a noise added interference matrix;

[0032] A second random mapping matrix is ​​generated according to the first random mapping matrix, the second random mapping matrix, the feature matching matrix, and the noise interference matrix.

[0033] Furthermore, the clustering operation includes:

[0034] Acquire a plurality of first cluster centers; wherein, initially, the first cluster centers are the initial cluster centers;

[0035] performing a clustering operation on the data to be clustered according to a plurality of the first clustering centers;

[0036] Calculate the total error value of each first cluster center, and divide the first cluster center with the smallest total error value into two second cluster centers;

[0037] Calculate the total number of centers of the first cluster center and the second cluster center;

[0038] When the total number of centers is less than the preset cluster center target value, the current first cluster center and the second cluster center are used as the first cluster center required for the next round of cluster division operation;

[0039] When the total number of centers is not less than a preset cluster center target value, the cluster division operation is terminated.

[0040] Another embodiment of the present invention provides a high-dimensional data clustering device based on random mapping, comprising:

[0041] A data acquisition module is used to acquire a high-dimensional data set consisting of several features and several preset dependent variables;

[0042] A data dimension reduction module, used to determine a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generate dimension reduction data according to the number of relevant features;

[0043] A first matrix generation module, used to generate a first random mapping matrix according to the number of relevant features in the dimension reduction data;

[0044] Wherein, the elements in the first random mapping matrix are all independent and identically distributed Gaussian random variables;

[0045] A transformation matrix generation module, used to perform random mapping transformation on the dimension reduction data according to the first random mapping matrix to generate a feature transformation matrix;

[0046] A second matrix generation module, used to generate a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix;

[0047] A random mapping module, used to perform random mapping transformation on the dimension reduction data using the second random mapping matrix to generate data to be clustered;

[0048] The data clustering module is used to take the data to be clustered as an initial cluster center and then repeat the clustering operation. When the number of cluster centers generated by the clustering operation is not less than a preset cluster center target value, the clustering result of the high-dimensional data set is generated.

[0049] Furthermore, the data dimension reduction module determines a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generates dimension reduction data according to the number of relevant features, including:

[0050] Using a symmetric uncertainty formula, calculate the correlation score between each dependent variable and each feature in the high-dimensional data set;

[0051] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a high-dimensional data clustering method based on random mapping as described in any one of the embodiments is implemented.

[0052] Another embodiment of the present invention provides a storage medium, characterized in that the storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute a high-dimensional data clustering method based on random mapping as described in any one of the above embodiments.

[0053] The following beneficial effects are achieved by implementing the present invention:

[0054] The present invention discloses a high-dimensional data clustering method, device, terminal device and storage medium based on random mapping. The method performs data dimensionality reduction operation on high-dimensional data according to preset dependent variables, retains only the features related to the dependent variables in the high-dimensional data, removes redundant features unrelated to the dependent variables, and greatly reduces the influence of redundant features on subsequent data clustering. Subsequently, before clustering, a random mapping matrix is ​​used to perform a random transformation on the reduced-dimensional data to achieve the effect of protecting the reduced-dimensional data. Therefore, the present invention avoids the interference of redundant features and the problem of local optimality of clustering by removing redundant features, and further protects privacy data through a random mapping matrix. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flowchart of a high-dimensional data clustering method based on random mapping provided by an embodiment of the present invention.

[0056] Figure 2 It is a structural schematic diagram of a high-dimensional data clustering device based on random mapping provided by an embodiment of the present invention.

[0057] Figure 3 It is a schematic diagram of randomly mapping dimensionality reduction data provided by an embodiment of the present invention.

[0058] Figure 4 It is a histogram of the number of cluster centers and cluster evaluation indicators provided by an embodiment of the present invention.

[0059] Figure 5 It is a flowchart of a privacy protection clustering process provided by an embodiment of the present invention.

[0060] Figure 6 This is a comparison chart of the anonymity rates of high-dimensional data aggregation results of different quantity-sensitive attributes provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0063] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined.

[0064] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0065] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0066] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0067] In the description of the embodiments of the present application, unless otherwise clearly specified and limited, technical terms such as "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0068] See also Figure 1 , is a flow chart of a high-dimensional data clustering method based on random mapping provided by an embodiment of the present invention, comprising:

[0069] S1. Obtain a high-dimensional data set consisting of several features and several preset dependent variables;

[0070] In a preferred embodiment of the present invention, high-dimensional data generally refers to high-dimensional and multivariate data, which contains data features that are different from two-dimensional and three-dimensional spatial data. Specifically, high-dimensional data refers to data with a very high dimension, even much larger than the number of samples. Data is very sparse in space, and the sample size is always very small compared to the dimension of the space.

[0071] In organizations such as medical institutions and educational institutions, a large amount of high-dimensional data is generated every day. In this embodiment, taking the data of medical institutions as an example, height, weight, blood type, etc. in the high-dimensional data set are the features of the high-dimensional data set. If the high-dimensional data is clustered to explore the relationship between height, weight and the incidence of diabetes, then the incidence of diabetes is the preset dependent variable.

[0072] S2. determining a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generating reduced-dimensionality data according to the number of relevant features;

[0073] In a preferred embodiment of the present invention, Figure 5 As shown in the figure, by reducing the dimensionality of high-dimensional data according to the dependent variable, relevant features with high correlation with the dependent variable can be determined, and some insignificant features can be removed, so that the subsequent clustering effect is better, the privacy of more valuable data can be protected, and more computing resources can be saved.

[0074] Preferably, determining a number of relevant features in the high-dimensional data set based on the number of dependent variables, and generating dimensionality reduction data based on the number of relevant features, includes:

[0075] S21, using a symmetric uncertainty formula to calculate the correlation score between each dependent variable and each feature in the high-dimensional data set;

[0076] S22. According to the correlation score, determine a plurality of initial redundant features and a plurality of initial relevant features. In a preferred embodiment of the present invention, the correlation score between each dependent variable and each feature in the high-dimensional data set is calculated by the following formula:

[0077]

[0078] Among them, I(X|Y) represents the pairwise information gain between the dependent variable Y and the feature X. G(X) represents the information entropy of the feature X, and G(Y) represents the information entropy of the dependent variable Y.

[0079] It can be understood that the range of the correlation score is [0,1]. If the correlation score is 1, it can be determined that feature X is correlated with dependent variable Y. If the correlation score is 0, it can be determined that feature X is irrelevant to dependent variable Y.

[0080] Preferably, after determining a plurality of initial redundant features and a plurality of initial relevant features, the method further comprises:

[0081] S221, normalizing the initial data under each initial related feature to generate normalized data under each initial related feature;

[0082] S222, plotting the normal distribution function of each initial correlation feature according to the normalized data under each initial correlation feature, and calculating the standard deviation of each state distribution function;

[0083] S223, obtaining preset data screening rules;

[0084] S224, when it is determined that the data screening rule is a moderate rule, deleting the normalized data that is greater than two times the standard deviation;

[0085] S225. When it is determined that the data screening rule is a strict rule, the normalized data that is larger than the standard deviation is deleted.

[0086] In a preferred embodiment of the present invention, in order to identify low-frequency data, first, normalization is used to refine the data under each feature; second, according to the characteristics of the standard normal distribution function, a normal distribution function is drawn for each column of the data; finally, strict rules or moderate rules are used to policy-based delete sensitive data.

[0087] Specifically, based on the principle of low duplication rate and greater possibility of privacy violation, in moderate mode, records with a difference exceeding two standard deviations are deleted when information is released. In strict mode, records with a difference equal to or greater than the standard deviation are deleted when information is released.

[0088] S23, using the Pearson correlation coefficient formula to calculate the Pearson coefficient between each dependent variable and each initial correlation feature;

[0089] S24: taking initial relevant features that are less than a preset coefficient threshold as redundant features, taking initial relevant features that are not less than the preset coefficient threshold as relevant features, and generating dimension reduction data according to the relevant features.

[0090] In a preferred embodiment of the present invention, for feature X and dependent variable Y, the Pearson correlation coefficient calculation formula between them is:

[0091]

[0092] It is understandable that the Pearson correlation coefficient is a tool for measuring the correlation between two random variables. According to the Pearson coefficient of each initial correlation feature, irrelevant redundant features can be further screened out. Therefore, in the data dimensionality reduction stage, this embodiment greatly removes redundant features through symmetric uncertainty and Pearson coefficient, greatly reducing the impact of data dimension on the privacy protection algorithm.

[0093] S3, generating a first random mapping matrix according to the number of relevant features in the dimension reduction data;

[0094] Wherein, the elements in the first random mapping matrix are all independent and identically distributed Gaussian random variables;

[0095] In a preferred embodiment of the present invention, when the number of relevant features in the dimension-reduced data is n, an independent and identically distributed n×n-dimensional random matrix R is generated, that is, R =[r 1 ,r 2 ,...,rn],ri(1≤i≤n) is an n-dimensional column vector whose elements are independent and identically distributed Gaussian random variables, that is, r ij : (1≤i,j≤n).

[0096] S4. Performing random mapping transformation on the dimension reduction data according to the first random mapping matrix to generate a feature transformation matrix;

[0097] Preferably, the feature transformation matrix includes: a feature matching matrix and a noise interference matrix to be added;

[0098] The step of performing random mapping transformation on the dimension reduction data according to the first random mapping matrix to generate a feature transformation matrix includes:

[0099] S41, orthogonalizing the first random mapping matrix, and then splitting the orthogonalized first random mapping matrix into a first random mapping sub-matrix and a second random mapping sub-matrix;

[0100] S42: Use the first random mapping sub-matrix and the second random mapping sub-matrix to perform random mapping transformation on the dimension reduction data to generate a feature matching matrix and a noise interference matrix to be added.

[0101] In a preferred embodiment of the present invention, in order to maintain the maximum similarity between the mapped vector and the original vector, the first random mapping matrix R is Gram-Schmidt orthogonalized, and the first random mapping matrix R is evenly divided to obtain mutually independent first random mapping sub-matrices R 1 and the second random mapping submatrix R 2 ,in

[0102] Then use the first random mapping submatrix R 1 and the second random mapping submatrix R 2 Perform random mapping transformation on the reduced-dimensional data to generate the feature transformation matrix y, which is then matched by the feature matching matrix y 1 and the interference matrix y to be added 2 The composition of which y 1 =R 1 T x,y 2 =R 2 T x, are all m×n dimensional features.

[0103] S5. Generate a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix;

[0104] Preferably, generating a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix includes:

[0105] S51, performing a noise adding operation on the interference matrix to be noised to generate a noise added interference matrix;

[0106] S52: Generate a second random mapping matrix according to the first random mapping matrix, the second random mapping matrix, the feature matching matrix, and the noise interference matrix.

[0107] In a preferred embodiment of the present invention, in the interference matrix y to be added with noise 2 Add random uniformly distributed m×n dimensional noise N s , get the noise interference matrix y 2ε =y 2 +N s .

[0108] Further, such as Figure 3 As shown, using the sub-random mapping matrix R 1 , R 2Cross-fusion transformation features to get y ε =R 2 y 1 +R 1 y 2ε , save it as the second random mapping matrix T.

[0109] S6. Using the second random mapping matrix to perform random mapping transformation on the dimension reduction data to generate data to be clustered;

[0110] S7. Taking the data to be clustered as an initial cluster center, and then repeating the clustering operation. When the number of cluster centers generated by the clustering operation is not less than a preset cluster center target value, a clustering result of the high-dimensional data set is generated.

[0111] Preferably, the clustering operation includes:

[0112] S71, obtaining a plurality of first cluster centers; wherein, initially, the first cluster centers are the initial cluster centers;

[0113] S72, performing a clustering operation on the data to be clustered according to a plurality of the first clustering centers;

[0114] S73, calculating the total error value of each first cluster center, and dividing the first cluster center with the smallest total error value into two second cluster centers;

[0115] S74, calculating the total number of centers of the first cluster center and the second cluster center;

[0116] S75: when the total number of centers is less than a preset cluster center target value, using the current first cluster center and the second cluster center as the first cluster center required for the next round of cluster division operation;

[0117] S76: When the total number of centers is not less than a preset cluster center target value, the cluster division operation is terminated.

[0118] In a preferred embodiment of the present invention, the k-means algorithm is used to cluster the clustered data. Specifically, the dimension reduction data is regarded as a cluster (cluster center) as a whole, and the k-means algorithm is used to cluster the clustered data, and then the total error of each cluster is calculated. When it is determined that the current number of clusters is less than the preset cluster center target value k, the cluster with the smallest error is selected for division until k clusters are reached. Figure 6 As shown, after the high-dimensional data with different quantitative sensitive attributes are aggregated through the high-dimensional data clustering method disclosed in this embodiment, the anonymity rate comparison chart of the aggregation results shows that the high-dimensional data clustering method disclosed in this embodiment can effectively protect the sensitive data in the high-dimensional data.

[0119] It should be noted that the goal of clustering is to improve the quality of clusters while keeping the number of clusters unchanged. Before clustering the data to be clustered, this embodiment first uses the following algorithm to obtain Figure 4 The results shown are then used to determine the final cluster center target value. Specifically, first, within the possible range of the number of clusters, different numbers of clusters are used to run the K-means algorithm to obtain the corresponding clustering results; then, a clustering validity function is constructed with the number of clusters k as the independent variable to evaluate the clustering results of the K-means, and the optimal number of clusters is determined by optimizing the clustering validity function.

[0120] This embodiment provides a high-dimensional data clustering method based on random mapping. By performing data dimensionality reduction operations on high-dimensional data according to preset dependent variables, only the features related to the dependent variables in the high-dimensional data are retained, and redundant features unrelated to the dependent variables are removed, which greatly reduces the impact of redundant features on subsequent data clustering. Subsequently, before clustering, a random mapping matrix is ​​used to perform a random transformation on the reduced-dimensional data to achieve the effect of protecting the reduced-dimensional data. Therefore, the present invention avoids the interference of redundant features and the problem of local optimality of clustering by removing redundant features, and further protects privacy data through a random mapping matrix.

[0121] See also Figure 2 , is a schematic diagram of a high-dimensional data clustering device based on random mapping provided by an embodiment of the present invention, comprising:

[0122] A data acquisition module is used to acquire a high-dimensional data set consisting of several features and several preset dependent variables;

[0123] A data dimension reduction module, used to determine a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generate dimension reduction data according to the number of relevant features;

[0124] A first matrix generation module, used to generate a first random mapping matrix according to the number of relevant features in the dimension reduction data;

[0125] Wherein, the elements in the first random mapping matrix are all independent and identically distributed Gaussian random variables;

[0126] A transformation matrix generation module, used to perform random mapping transformation on the dimension reduction data according to the first random mapping matrix to generate a feature transformation matrix;

[0127] A second matrix generation module, used to generate a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix;

[0128] A random mapping module, used to perform random mapping transformation on the dimension reduction data using the second random mapping matrix to generate data to be clustered;

[0129] The data clustering module is used to take the data to be clustered as an initial cluster center and then repeat the clustering operation. When the number of cluster centers generated by the clustering operation is not less than a preset cluster center target value, the clustering result of the high-dimensional data set is generated.

[0130] Preferably, the data dimension reduction module determines a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generates dimension reduction data according to the number of relevant features, including:

[0131] Using a symmetric uncertainty formula, calculate the correlation score between each dependent variable and each feature in the high-dimensional data set;

[0132] Determining a number of initial redundant features and a number of initial relevant features according to the relevance score;

[0133] The Pearson correlation coefficient formula is used to calculate the Pearson coefficient between each dependent variable and each initial correlation feature;

[0134] The initial relevant features that are less than a preset coefficient threshold are used as redundant features, and the initial relevant features that are not less than the preset coefficient threshold are used as relevant features, and the dimension reduction data are generated according to the relevant features.

[0135] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0136] Those skilled in the art can clearly understand that for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0137] Another preferred embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements a high-dimensional data clustering method based on random mapping as described in any one of the above embodiments.

[0138] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0139] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0140] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0141] Another preferred embodiment of the present invention provides a storage medium, the storage medium is a computer-readable storage medium, the computer program is stored in the computer-readable storage medium, and the computer program, when executed by the processor, can implement the steps of the above-mentioned various method embodiments. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0142] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A high-dimensional data clustering method based on random mapping, characterized in that: include: Obtain a high-dimensional data set consisting of several features and several preset dependent variables; Determining a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generating reduced-dimensionality data according to the number of relevant features; generating a first random mapping matrix according to the number of relevant features in the dimensionally reduced data; Wherein, the elements in the first random mapping matrix are all independent and identically distributed Gaussian random variables; According to the first random mapping matrix, performing random mapping transformation on the dimension reduction data to generate a feature transformation matrix; Generate a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix; Using the second random mapping matrix to perform random mapping transformation on the dimension reduction data to generate data to be clustered; The data to be clustered is taken as an initial cluster center, and the clustering operation is then repeated. When the number of cluster centers generated by the clustering operation is not less than a preset cluster center target value, a clustering result of the high-dimensional data set is generated.

2. A high-dimensional data clustering method based on random mapping as claimed in claim 1, characterized in that: Determining a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generating reduced-dimensionality data according to the number of relevant features, includes: Using a symmetric uncertainty formula, calculate the correlation score between each dependent variable and each feature in the high-dimensional data set; Determining a number of initial relevant features according to the relevance score; The Pearson correlation coefficient formula is used to calculate the Pearson coefficient between each dependent variable and each initial correlation feature; The initial relevant features that are less than a preset coefficient threshold are used as redundant features, and the initial relevant features that are not less than the preset coefficient threshold are used as relevant features, and the dimension reduction data are generated according to the relevant features.

3. A high-dimensional data clustering method based on random mapping as claimed in claim 2, characterized in that: After determining a number of initial redundant features and a number of initial relevant features, it also includes: Normalizing the initial data under each initial relevant feature to generate normalized data under each initial relevant feature; According to the normalized data under each initial correlation feature, the normal distribution function of each initial correlation feature is plotted, and the standard deviation of each state distribution function is calculated; Get the preset data filtering rules; When it is determined that the data screening rule is a moderate rule, the normalized data that is greater than two times the standard deviation is deleted; When the data screening rule is determined to be a strict rule, the normalized data that is larger than the standard deviation is deleted.

4. A high-dimensional data clustering method based on random mapping as claimed in claim 1, characterized in that: The feature transformation matrix includes: a feature matching matrix and a noise interference matrix to be added; The step of performing random mapping transformation on the dimension reduction data according to the first random mapping matrix to generate a feature transformation matrix includes: Orthogonalizing the first random mapping matrix, and then splitting the orthogonalized first random mapping matrix into a first random mapping sub-matrix and a second random mapping sub-matrix; The first random mapping sub-matrix and the second random mapping sub-matrix are used to perform random mapping transformation on the dimension reduction data to generate a feature matching matrix and an interference matrix to be noisy.

5. A high-dimensional data clustering method based on random mapping as claimed in claim 4, characterized in that: The step of generating a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix includes: Performing a noise adding operation on the interference matrix to be noised to generate a noise added interference matrix; A second random mapping matrix is ​​generated according to the first random mapping matrix, the second random mapping matrix, the feature matching matrix, and the noise interference matrix.

6. A high-dimensional data clustering method based on random mapping as claimed in claim 1, characterized in that: The clustering operation includes: Acquire a plurality of first cluster centers; wherein, initially, the first cluster centers are the initial cluster centers; performing a clustering operation on the data to be clustered according to a plurality of the first clustering centers; Calculate the total error value of each first cluster center, and divide the first cluster center with the smallest total error value into two second cluster centers; Calculate the total number of centers of the first cluster center and the second cluster center; When the total number of centers is less than the preset cluster center target value, the current first cluster center and the second cluster center are used as the first cluster center required for the next round of cluster division operation; When the total number of centers is not less than a preset cluster center target value, the cluster division operation is terminated.

7. A high-dimensional data clustering device based on random mapping, characterized in that: include: A data acquisition module is used to acquire a high-dimensional data set consisting of several features and several preset dependent variables; A data dimension reduction module, used to determine a number of relevant features in the high-dimensional data set according to the number of dependent variables, and generate dimension reduction data according to the number of relevant features; A first matrix generation module, used to generate a first random mapping matrix according to the number of relevant features in the dimension reduction data; Wherein, the elements in the first random mapping matrix are all independent and identically distributed Gaussian random variables; A transformation matrix generation module, used to perform random mapping transformation on the dimension reduction data according to the first random mapping matrix to generate a feature transformation matrix; A second matrix generation module, used to generate a second random mapping matrix according to the feature transformation matrix and the first random mapping matrix; A random mapping module, used to perform random mapping transformation on the dimension reduction data using the second random mapping matrix to generate data to be clustered; The data clustering module is used to take the data to be clustered as an initial cluster center and then repeat the clustering operation. When the number of cluster centers generated by the clustering operation is not less than a preset cluster center target value, the clustering result of the high-dimensional data set is generated.

8. A high-dimensional data clustering device based on random mapping as claimed in claim 7, characterized in that: The data dimension reduction module determines a number of related features in the high-dimensional data set according to the number of dependent variables, and generates dimension reduction data according to the number of related features, including: Using a symmetric uncertainty formula, calculate the correlation score between each dependent variable and each feature in the high-dimensional data set; Determining a number of initial redundant features and a number of initial relevant features according to the relevance score; The Pearson correlation coefficient formula is used to calculate the Pearson coefficient between each dependent variable and each initial correlation feature; The initial relevant features that are less than a preset coefficient threshold are used as redundant features, and the initial relevant features that are not less than the preset coefficient threshold are used as relevant features, and the dimension reduction data are generated according to the relevant features.

9. A terminal device, characterized in that: It comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a high-dimensional data clustering method based on random mapping as described in any one of claims 1 to 6 is implemented.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute a high-dimensional data clustering method based on random mapping as described in any one of claims 1 to 6.

Citation Information

Patent Citations

  • High-dimensional private data publishing method based on Bayesian network attribute clustering analysis technology

    CN112131604A

  • Construction method of correlation data transaction framework oriented to high-dimensional private data

    CN115630964A

  • Differential privacy data desensitization method based on time sequence random mapping and related device

    CN116011011A

  • Privacy-preserving fast approximate k-means clustering with hamming vectors

    US20210019556A1