Method for improving clustering performance

Abstract

The invention relates to the technical field of data mining, and discloses a method for improving clustering performance, which comprehensively applies technical means such as subspace learning, nonlinear processing, tensor nuclear norm constraint and consensus similarity learning. Through subspace learning, the method can reveal hidden structures and features in high-dimensional data, and effectively deals with dimension disasters. The nonlinear processing part optimizes nonlinear information in the data and reveals a deep nonlinear structure of the data. By using the tensor nuclear norm constraint, the method not only improves the performance of a clustering algorithm, but also shows a remarkable effect in practical application, for example, the recognition accuracy is improved in face recognition, analysts are helped to more accurately distinguish different consumer groups in the market analysis field and insight into consumer demands, and the market analysis efficiency is improved. Therefore, the accuracy and effectiveness of market analysis are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and in particular to a method for data processing. Background Art

[0002] Data mining is a hot topic in the field of artificial intelligence and database research. Data mining refers to the non-trivial process of revealing implicit, previously unknown and potentially valuable information from a large amount of data in a database. Data mining is a decision support process that is mainly based on artificial intelligence, machine learning, pattern recognition, statistics, databases, visualization technology, etc. It analyzes the data of enterprises in a highly automated manner, makes inductive inferences, and mines potential patterns to help decision makers adjust market strategies, reduce risks, and make correct decisions. In the future, with the continuous advancement of technology and the continuous expansion of application fields, data will become more and more numerous and complex. Data mining technology will play an important role in more fields. This application now proposes a method for improving clustering performance. Summary of the invention

[0003] In view of the deficiencies of the prior art, the present invention provides a method for improving clustering performance, which improves clustering performance by effectively utilizing nonlinear information in data and complementary information of multi-view data.

[0004] To achieve the above object, the present invention provides the following technical solution: a method for improving clustering performance, comprising the following steps:

[0005] S1, subspace learning, reveals the hidden structure and features in high-dimensional data, and effectively addresses the dimensionality curse problem;

[0006] S2, nonlinear processing, optimizing the nonlinear information hidden in the data and revealing the nonlinear structure of the data;

[0007] S3, tensor core norm constraint, captures complementary information between views and explores high-order correlations between different views;

[0008] S4, consensus similarity learning, can learn a consistent clustering indicator matrix for clustering.

[0009] Preferably, the subspace learning adopts a self-representation-based method to obtain a coefficient matrix that expresses the intrinsic relationship of the data.

[0010] Preferably, the nonlinear processing adopts a kernel method to transform the original data into a high-dimensional feature space in the Hilbert space through nonlinear mapping, which can reveal and utilize the nonlinear structure implicit in the data.

[0011] Preferably, the tensor nuclear norm constraint step is as follows:

[0012] S301, constructing coefficient matrices of different views into third-order tensors;

[0013] S302. Constrain the constructed third-order tensor using the tensor nuclear norm.

[0014] Preferably, the consensus similarity learning steps are as follows:

[0015] S401, construct a consensus Laplacian matrix using consensus similarity learning;

[0016] S402: learning a consistent clustering indicator matrix through a consensus Laplacian matrix.

[0017] Compared with the prior art, the present invention provides a method for improving clustering performance, which has the following beneficial effects:

[0018] 1. The clustering performance improvement method comprehensively uses subspace learning, nonlinear processing, tensor nuclear norm constraints, and consensus similarity learning to improve the efficiency and accuracy of clustering algorithms. Through subspace learning, this method can reveal the hidden structure and features in high-dimensional data and effectively deal with the dimensionality curse problem. The nonlinear processing part optimizes the nonlinear information in the data and reveals the deep nonlinear structure of the data. Using the tensor nuclear norm constraint, this method can explore and mine the high-order correlation between multi-view data, further enhancing the depth of data analysis. The introduction of consensus similarity learning constructs a consensus Laplace matrix to obtain a consistent clustering indicator matrix for more accurate clustering. The comprehensive application of these technologies not only improves the performance of clustering algorithms, but also shows significant effects in practical applications, such as improving the recognition accuracy in face recognition, and helping analysts to more accurately distinguish different consumer groups in the field of market analysis and gain insight into consumer needs, thereby significantly improving the accuracy and effectiveness of market analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The figure is a flow chart of clustering performance improvement of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] See also Figure 1 , a method for improving clustering performance, comprising the following steps:

[0022] S1, subspace learning, reveals the hidden structure and features in high-dimensional data, and effectively addresses the dimensionality curse problem;

[0023] S2, nonlinear processing, optimizing the nonlinear information hidden in the data and revealing the nonlinear structure of the data;

[0024] S3, tensor core norm constraint, captures complementary information between views and explores high-order correlations between different views;

[0025] S4, consensus similarity learning, can learn a consistent clustering indicator matrix for clustering.

[0026] Embodiment 1: A method for improving clustering performance, comprising the following steps:

[0027] S1, subspace learning, reveals the hidden structure and features in high-dimensional data, and effectively addresses the dimensionality curse problem;

[0028] S2, nonlinear processing, optimizing the nonlinear information hidden in the data and revealing the nonlinear structure of the data;

[0029] S3, tensor core norm constraint, captures complementary information between views and explores high-order correlations between different views;

[0030] S4, consensus similarity learning, can learn a consistent clustering indicator matrix for clustering.

[0031] Embodiment 2: A method for improving clustering performance according to embodiment 1, wherein subspace learning adopts a self-representation-based method to obtain a coefficient matrix that expresses the intrinsic relationship of data.

[0032] Embodiment 3: A method for improving clustering performance according to embodiment 1, wherein the nonlinear processing uses a kernel method to transform the original data into a high-dimensional feature space in the Hilbert space through nonlinear mapping, which can reveal and utilize the nonlinear structure implicit in the data.

[0033] Embodiment 4: According to a method for improving clustering performance in Embodiment 1, the steps of constraining the tensor nuclear norm are as follows:

[0034] S301, constructing coefficient matrices of different views into third-order tensors;

[0035] S302. Constrain the constructed third-order tensor using the tensor nuclear norm.

[0036] Embodiment 5: According to a method for improving clustering performance in Embodiment 1, the consensus similarity learning steps are as follows:

[0037] S401, construct a consensus Laplacian matrix using consensus similarity learning;

[0038] S402: learning a consistent clustering indicator matrix through a consensus Laplacian matrix.

[0039] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for improving clustering performance, characterized in that: The following steps are involved: S1, subspace learning, reveals the hidden structure and features in high-dimensional data, and effectively addresses the dimensionality curse problem; S2, nonlinear processing, optimizing the nonlinear information hidden in the data and revealing the nonlinear structure of the data; S3, tensor core norm constraint, captures complementary information between views and explores high-order correlations between different views; S4, consensus similarity learning, can learn a consistent clustering indicator matrix for clustering.

2. The method for improving clustering performance according to claim 1, characterized in that: The subspace learning adopts a self-representation-based method to obtain a coefficient matrix that expresses the intrinsic relationship of the data.

3. The method for improving clustering performance according to claim 1, characterized in that: The method based on self-representation is used to obtain the coefficient matrix expressing the intrinsic relationship of the data, which includes the following steps: dimensionality reduction and feature extraction are achieved by training two interconnected networks, one of which is called an encoder and the other is called a decoder. The encoder converts the input data into a form from the original size to a low-dimensional representation, and the decoder reconstructs the representation into the form of the input data. By establishing a residual relationship between the encoder and the decoder, the coefficient matrix of the intrinsic relationship of the data can be obtained using the autoencoder to obtain the coefficient matrix expressing the intrinsic relationship of the data.

4. The method for improving clustering performance according to claim 1, characterized in that: The nonlinear processing adopts a kernel method to transform the original data into a high-dimensional feature space in the Hilbert space through nonlinear mapping, which can reveal and utilize the nonlinear structure implicit in the data.

5. The method for improving clustering performance according to claim 1, characterized in that: The steps of tensor nuclear norm constraint are as follows: S301, constructing coefficient matrices of different views into third-order tensors; S302. Constrain the constructed third-order tensor using the tensor nuclear norm.

6. The method for improving clustering performance according to claim 1, characterized in that: The consensus similarity learning steps are as follows: S401, construct a consensus Laplacian matrix using consensus similarity learning; S402: learning a consistent clustering indicator matrix through a consensus Laplace matrix.

Images