Method and system for shadow set fuzzy clustering based on decision weight driving

Abstract

The invention relates to the technical field of data processing, and particularly discloses a shadow set fuzzy clustering method and system based on decision weight driving, and the method comprises the steps: carrying out the clustering initialization; different decision weight matrixes are generated by calculating a membership matrix and constructing a decision weight function so as to accurately measure the influence weights of different samples in different areas on the cluster center; constructing a fuzzy cluster shadow set: establishing an optimization principle, calculating a partition threshold value of each fuzzy cluster through a particle swarm optimization algorithm, and dividing each fuzzy cluster into a core region, a shadow region and an exclusion region; updating the clustering center: updating the clustering center based on the contribution difference of each region sample to ensure that the contribution of the sample of the core region to the updating is maximum, and meanwhile, excluding the region sample is brought into calculation to prevent the clustering center from falling into local optimum; iterating and stopping when a stopping condition is met. According to the method, the advantages of the decision weighting function conforming to fuzzy cluster distribution and the shadow set are effectively combined, and the accuracy and robustness of fuzzy clustering are improved.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and specifically discloses a shadow set fuzzy clustering method and system based on decision weight drive. Background Art

[0002] In the fuzzy clustering method, samples are fuzzily divided by fuzzy membership, so that samples can belong to multiple clusters at the same time. This fuzzy division can better handle the uncertainty and ambiguity in the data. However, the traditional fuzzy clustering method has a relatively single weight processing method for all samples in the clustering process, and fails to fully consider the differences in the degree of influence of different samples on the update of cluster centers.

[0003] In contrast, the shadow set fuzzy clustering method introduces a mechanism to divide the fuzzy cluster into three areas: core area, shadow area and exclusion area. The advantage of the shadow set is that it can dynamically adjust the weights of different areas on the cluster center update according to the degree of membership of the sample to the cluster center, thereby effectively suppressing the adverse effects of outliers and noise data on the cluster center update and improving the robustness and accuracy of the clustering results. Specifically, the samples in the core area have the greatest impact on the cluster center, followed by the shadow area, while the samples in the exclusion area have the least impact on the update of the cluster center.

[0004] However, the traditional shadow set fuzzy clustering method processes samples in a single region. Current methods usually regard samples in the same region as having the same weight, which ignores the difference in the impact of different samples in the same region on the cluster center.

[0005] In view of this, the inventor provides a shadow set fuzzy clustering method and system driven by decision weights, so as to improve the traditional shadow set fuzzy clustering method. Summary of the invention

[0006] The purpose of the present invention is to improve the traditional shadow set fuzzy clustering method which treats the weights of samples in the same area equally. The traditional method usually regards the samples in the same area as having the same weights, while ignoring the differences in the impact of different samples on the cluster center.

[0007] In order to achieve the above object, the basic scheme of the present invention provides a shadow set fuzzy clustering method based on decision weight driving, comprising the following steps: Step S1: Cluster initialization: Initialize the cluster center and set the parameters required in the clustering process, including the objective function difference threshold, fuzzy factor, and maximum number of iterations; Step S2: Calculate the decision weight matrix: Calculate the membership matrix and use the decision weight function to map the membership matrix into a new decision weight matrix; Step S3: constructing a fuzzy cluster shadow set: establishing an optimization principle, using a particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster, thereby constructing a fuzzy cluster shadow set; Step S4: Update the cluster center: construct a cluster center update formula based on the contribution of different partitions in the fuzzy cluster shadow set to the cluster center; Step S5: Iterate step S2 to step S4 until clustering ends after reaching the iteration stop condition, wherein the iteration stop condition is: comparing the difference of the objective function before and after the update and stopping when the objective function difference is less than the difference threshold or when reaching the maximum number of iterations.

[0008] Further, in step S1, cluster initialization includes the following steps: Step S11: Initialize cluster centers: For a data set X={x1,x2,…,x n} Randomly select k cluster centers or use K-means to initialize k fuzzy cluster centers; Step S12: Setting the parameters required in the clustering process, including the objective function difference threshold ε, the fuzzy factor m, and the maximum number of iterations T.

[0009] Further, in step S2, calculating the decision weight matrix specifically includes the following steps: Step S21: Calculate the membership matrix: calculate the membership of each sample to each cluster center; Step S22: constructing a decision weight function; Step S23: Calculate the decision weight matrix.

[0010] Further, in step S22, the constructed decision weight function specifically includes: Step S221: Constructing a segmented decision weight function that conforms to the fuzzy cluster distribution , expressed as: ; in, Represents the segment decision weight function The next j-th sample x j Belongs to the i-th cluster center c i The decision weight value, u ij represents the jth sample x j Belongs to the i-th cluster center c i The membership degree is 1≤j≤n, 1≤i≤k, w n , w m and w p They are the membership intervals [0,α i ), [α i ,β i ] and (β i,1], the decision weight function is specifically expressed as: ; ; ; Among them, (α i ,β i ) is a pair of thresholds, satisfying 0<α i <β i <1, parameters , , , is the shape parameter. The parameter σ is used to determine the key point (θ3, 0.5), which is used to adjust the curvature of the decision weight function. In addition, θ1=1 and θ2=0 are set. Step S222: Construct an S-type decision weight function that conforms to the fuzzy cluster distribution , expressed as: ; in, Represents the segment decision weight function The next j-th sample x j Belongs to the i-th cluster center c i The decision weight value, u ij represents the jth sample x j Belongs to the i-th cluster center c i The membership degree is 1≤j≤n, 1≤i≤k.

[0011] Further, in step S3, the specific steps of constructing the fuzzy cluster shadow set are: Step S31: Establishing optimization principles: setting three mapping sets of fuzzy sets to establish optimization principles for constructing fuzzy cluster shadow sets; Step S32: using a particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster; Step S33: construct a fuzzy cluster shadow set, each fuzzy cluster is represented as c i ={Inc(c i ),shd(c i ),Exl(c i )}.

[0012] Further, in step S31, the specific steps of establishing the optimization principle are: Step S311: Determine the three-value set of fuzzy set mapping: for any fuzzy cluster center c i , the decision weight values ​​of all samples are regarded as a fuzzy set, represented by wi ={w ij |j=1,2,…,n}, construct fuzzy set w i The three mapping sets on : ; Among them, (α i ,β i ) is a pair of thresholds, satisfying 0<α i <β i <1,w ij Represents sample x j Belongs to cluster center c i The decision weight value of Step S312: Calculate the partition threshold (α i ,β i ), using the minimum error optimization principle: ; Among them, L is the target optimization function, w ij Represents sample x j Belongs to cluster center c i The decision weight value, and w ij ∈[0,1], which can be specifically expressed as and .

[0013] Further, in step S4, the specific steps of updating the cluster center are: For any fuzzy cluster c i , whose blurred shadow is represented by c i ={Inc(c i ),shd(c i ),Exl(c i )}, according to the difference in contribution of samples in different areas of the fuzzy cluster shadow set to the update of the cluster center, the following cluster center update formula is constructed based on the decision weight matrix: ; Among them, card is the cardinality, X is the data set, and X={x1,x2,…,x n},(α i ,β i ) is the partition threshold of each fuzzy cluster, w ij Represents sample x j Belongs to cluster center c i The decision weight value of .

[0014] Further, in step S5, the iteration stop condition is specifically: (1) Compare the difference between the objective functions J(t) and J(t-1) updated at the t-th iteration and the t-1-th iteration. If J(t)-J(t-1)≤ε; The iteration stops, Among them, ε is the objective function difference threshold; Among them, the objective function of each iteration is expressed as follows: ; Where W represents the decision weight value matrix induced by the membership matrix, C represents the cluster center set, and C=(c1,c2,…,c k ), m is the fuzzification parameter and m≥1, d ij Represents sample x j With cluster center c i Euclidean distance, 1≤j≤n, 1≤i≤k, w ij Represents sample x j Belongs to cluster center c i The decision weight value, and w ij ∈[0,1]; (2) Or the process stops when the number of iterations reaches the maximum number of iterations T.

[0015] The basic scheme of the present invention also provides a shadow set fuzzy clustering system based on decision weight driving to implement the shadow set fuzzy clustering method based on decision weight driving as described above, comprising: Initialization module: used to initialize the cluster center and set the parameters required in the clustering process; Decision weight matrix calculation module: used to calculate the membership matrix and use the decision weight function to map the membership matrix into a new decision weight matrix; Constructing fuzzy cluster shadow set module: used to establish optimization principles and use particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster, so as to construct fuzzy cluster shadow set; Cluster center update module: used to construct the cluster center update formula according to the contribution of different partitions in the fuzzy cluster shadow set to the cluster center; Check iterative stop condition module: used for iterative calculation, compares the difference of objective function before and after update and stops when the difference of objective function is less than the difference threshold or when the maximum number of iterations is reached.

[0016] The principle and effect of this scheme are: 1. Compared with the prior art, the present invention introduces a more sophisticated weight allocation mechanism to improve the traditional shadow set fuzzy clustering method. For a fuzzy cluster, the decision weights of the cluster center are differentiated according to different samples in the same area, which overcomes the limitation of the traditional method of uniformly assigning the same weight to samples in the same area. Specifically, samples with larger decision weights contribute more significantly to the update of the cluster center, while samples with smaller decision weights have relatively weaker effects on the update of the cluster center. This improved strategy effectively solves the problem of the traditional shadow set fuzzy clustering method treating sample weights equally in a single area, and fully considers the differences in the degree of influence of different samples on the cluster center in the same area, thereby improving the accuracy and robustness of the clustering results.

[0017] 2. Compared with the prior art, the present invention enhances the robustness to noise and outliers. In the clustering process, by dividing the fuzzy cluster into three parts: core area, shadow area and exclusion area, different weights are assigned to samples in different areas, where the core area has the greatest impact on the weight of the cluster center, while the exclusion area has the least impact on the weight. This mechanism reduces the adverse effects of outliers and noise on the update of the cluster center, and effectively improves the robustness of clustering.

[0018] 3. Compared with the prior art, the present invention has enhanced ability to adapt to complex data sets. By dynamically adjusting the weights of samples to cluster centers, the clusters generated by clustering results can better adapt to data sets with complex boundaries. This method shows strong adaptability when processing data sets with complex data distribution or fuzzy boundaries, which helps to obtain better clustering results in data environments with high uncertainty. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 A flowchart of a shadow set fuzzy clustering method based on decision weight driving proposed in an embodiment of the present application is shown; Figure 2 A schematic diagram of a segmented decision weight function proposed in an embodiment of the present application is shown; Figure 3 A schematic diagram of an S-type decision weight function proposed in an embodiment of the present application is shown; Figure 4 A schematic diagram of a fuzzy cluster shadow set derived from a segmented decision weight function proposed in an embodiment of the present application is shown; Figure 5A schematic diagram of a fuzzy cluster shadow set derived from an S-type decision weight function proposed in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0021] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0022] A shadow set fuzzy clustering method based on decision weight drive is implemented, for example Figure 1 As shown, the following steps are included: Step S1: Cluster initialization: Initialize the cluster center and set the parameters required in the clustering process, including the objective function difference threshold, fuzzy factor, and maximum number of iterations.

[0023] Specifically, cluster initialization includes: n} Randomly select k cluster centers or use K-means to initialize k cluster centers C=(c1,c2,…,c k ). Set the parameters required in the clustering process, including the objective function difference threshold ε, the fuzzy factor m, and the maximum number of iterations T.

[0024] Step S2: Calculate the decision weight matrix: Calculate the membership matrix and use the decision weight function to map the membership matrix into a new decision weight matrix.

[0025] Specifically, the calculation of the decision weight matrix includes the following steps: Step S21: Calculate the membership matrix U k×n :Calculate the jth sample x j Belongs to the i-th cluster center c i The membership degree u ij , where 1≤j≤n, 1≤i≤k, the calculation formula is as follows: ; Among them, u ij Represents sample x j Belongs to cluster center c i The membership degree of u ij ∈[0,1],d ij and d lj Respectively represent samples x j With cluster center c i and other cluster centers c l The Euclidean distance of , 1≤l≤k, and l≠i, k is the number of fuzzy clusters.

[0026] Step S22: Constructing a decision weight function ; Step S221: Constructing a segmented decision weight function that conforms to the fuzzy cluster distribution ,like Figure 2 As shown, the expression is: ; in, Represents the segment decision weight function The next j-th sample x j Belongs to the i-th cluster center c i The decision weight value, u ij represents the jth sample x j Belongs to the i-th cluster center c i The membership degree is 1≤j≤n, 1≤i≤k, w n , w m and w p They are the membership intervals [0,α i ), [α i ,β i ] and (β i ,1], the decision weight function is specifically expressed as: ; ; ; Among them, (α i ,β i ) is a pair of thresholds, satisfying 0<α i <β i <1, parameters , , , is the shape parameter. The parameter σ is used to determine the key point (θ3, 0.5), which is used to adjust the curvature of the decision weight function. In addition, θ1=1 and θ2=0 are set. Step S222: Construct an S-type decision weight function that conforms to the fuzzy cluster distribution ,like Figure 3 As shown, it is expressed as: ; in, Represents the segment decision weight function The next j-th sample x j Belongs to the i-th cluster center c i The decision weight value, u ij represents the jth sample x j Belongs to the i-th cluster center c iThe membership degree is 1≤j≤n, 1≤i≤k.

[0027] Step S23: Calculate the decision weight matrix W k×n ; ; Among them, W k×n is the decision weight matrix, U k×n is the membership matrix, It's U k×n To W k×n The decision weight mapping function can be specifically expressed as or .

[0028] Step S3: constructing fuzzy cluster shadow sets: establishing optimization principles, using particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster, and thus constructing fuzzy cluster shadow sets.

[0029] Specifically, the steps include: Step S31: Establishing optimization principles: setting three mapping sets of fuzzy sets to establish optimization principles for constructing fuzzy cluster shadow sets; Step S311: Determine the three-value set of fuzzy set mapping: membership matrix U k×n It refers to the membership of n samples to the cluster center of k, and the decision weight function maps the membership value to the decision weight value. Therefore, for any fuzzy cluster center c i , the decision weight values ​​of all samples are regarded as a fuzzy set, represented by w i ={w ij |j=1,2,…,n}, construct fuzzy set w i The three mapping sets on : ; From the above formula, we can see that the threshold (α i ,β i ) is the definite fuzzy set w i The key of the three-branch mapping, and satisfies 0<α i <β i <1; Step S312: Calculate the partition threshold (α i ,β i ), using the minimum error optimization principle: ;

[0030] Among them, L is the target optimization function, w ij Represents sample x j Belongs to cluster center c i The decision weight value, and w ij∈[0,1], which can be specifically expressed as and .

[0031] Step S32: Use the particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster, specifically: The particle swarm optimization algorithm is used to calculate the partition threshold (α i ,β i ), the objective function of the minimum error optimization principle is regarded as the fitness function of the particle swarm optimization algorithm, and the constraint condition is 0<α i <β i <1, and obtain the partition threshold (α i ,β i ), thus the fuzzy cluster is represented as the fuzzy cluster shadow set c i ={Inc(c i ),shd(c i ),Exl(c i )},in Inc(c i )={x j ∈X|β i <w ij ≤1}; shd(c i )={x j ∈X|α i ≤w ij ≤β i}; Exl(c i )={x j ∈X│0≤w ij <α i}; Step S33: construct a fuzzy cluster shadow set, denoted as c i ={Inc(c i ),shd(c i ),Exl(c i )}.

[0032] like Figure 4 Shown is a schematic diagram of the fuzzy cluster shadow set derived from the piecewise decision weight function.

[0033] like Figure 5 As shown, it is a schematic diagram of the fuzzy cluster shadow set derived from the S-type decision weight function.

[0034] Step S4: Update the cluster center: construct a cluster center update formula based on the contribution of different partitions in the fuzzy cluster shadow set to the cluster center.

[0035] Specifically: For any fuzzy cluster c i , whose blurred shadow is represented by c i ={Inc(c i ),shd(c i ),Exl(c i )}, according to the difference in contribution of samples in different areas of the fuzzy cluster shadow set to the update of the cluster center, the following cluster center update formula is constructed based on the decision weight matrix: ; Among them, card is the cardinality, X is the data set, and X={x1,x2,…,x n},(α i ,β i ) is the partition threshold of each fuzzy cluster, w ij Represents sample x j Belongs to cluster center c i The decision weight value of .

[0036] It is worth noting that the core area of ​​the fuzzy cluster Inc(c i ) must belong to this cluster, so the samples in this area have a greater weight for updating the cluster center and are set to 1; In the shadow area of ​​the fuzzy cluster shd(c i ) may belong to the cluster or not, so the weight of the samples in this area for updating the cluster center should be smaller than the weight of the samples in the core area, which is set to be the same as the sample weight in the fuzzy C-means clustering process, which is set to (w ij ) m ; In the fuzzy cluster exclusion area Exl(c i ) do not belong to this cluster, so the weight of the samples in this area for updating the cluster center should be smaller than the weight of the samples in the shadow area, so it is set to .

[0037] In addition, even if the samples in the excluded area do not belong to the cluster, they still participate in the update calculation of the cluster center, avoiding the cluster center falling into local search during the clustering iterative update.

[0038] Step S5: Iterate step S2 to step S4 until clustering ends after reaching the iteration stop condition, wherein the iteration stop condition is specifically: (1) Compare the difference between the objective functions J(t) and J(t-1) updated at the t-th iteration and the t-1-th iteration. If J(t)-J(t-1)≤ε; The iteration stops, Among them, ε is the objective function difference threshold; Among them, the objective function of each iteration is expressed as: ; Where W represents the decision weight value matrix induced by the membership matrix, C represents the cluster center set, and C=(c1,c2,…,c k ), m is the fuzzification parameter and m≥1, d ij Represents sample x j With cluster center c i Euclidean distance, 1≤j≤n, 1≤i≤k, w ij Represents sample x j Belongs to cluster center c i The decision weight value, and w ij ∈[0,1]; (2) Or the process stops when the number of iterations reaches the maximum number of iterations T.

[0039] In this embodiment, step S6 is also included: evaluation: using clustering indicators to evaluate clustering results.

[0040] Specifically, performance indicators such as XB, DB, PBM, Kn, etc. are used to evaluate the clustering results.

[0041] In accordance with the above technical solution, a shadow set fuzzy clustering method based on decision weight drive in this embodiment is experimented on 5 UCI machine learning datasets. The experimental datasets are randomly selected from the database of University of California, Irvine (https: / / archive.ics.uci.edu / datasets), and the basic information is shown in Table 1.

[0042] Table 1: UCI machine learning dataset

[0043] This embodiment has achieved some positive results in the process of research and development and use, and has indeed great advantages compared with the prior art. The following content is described in combination with the data of the test process and Table 2.

[0044] Table 2: Experimental results of different shadow set fuzzy clustering methods on the UCI dataset

[0045] Table 2 shows the experimental results of four shadow set fuzzy clustering methods on five UCI datasets with respect to the evaluation indicators XB, DB, PBM and KN. The best results are indicated in bold, and the suboptimal results are indicated in underline. As can be seen from the results, the marked results are mostly concentrated on W P SCM and W SSCM two shadow set fuzzy clustering methods based on decision weight drive. It can be seen that the clustering result of the clustering method proposed in the present invention has higher quality, so it can be proved that the clustering method proposed by the present invention through improving the shadow set fuzzy clustering process is effective.

[0046] The present invention also provides a shadow set fuzzy clustering system based on decision weight driving, comprising: Initialization module: used to initialize the cluster center and set the parameters required in the clustering process; Decision weight matrix calculation module: used to calculate the membership matrix and use the decision weight function to map the membership matrix into a new decision weight matrix; Constructing fuzzy cluster shadow set module: used to establish optimization principles and use particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster, so as to construct fuzzy cluster shadow set; Cluster center update module: used to consider the contribution of different partitions in the fuzzy cluster shadow set to the cluster center and construct the cluster center update formula; Iteration stop condition check module: used for iterative calculation, compares the difference between the objective function before and after the update and stops when the objective function difference is less than the difference threshold or when the maximum number of iterations is reached; Evaluation module: Use performance indicators such as XB, DB, PBM, Kn, etc. to evaluate the clustering results.

[0047] The present invention also provides a computer device including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the shadow set fuzzy clustering method driven by decision weights.

[0048] The present invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the shadow set fuzzy clustering method driven by decision weights.

[0049] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A shadow set fuzzy clustering method based on decision weight drive, characterized in that: The steps include: Step S1: Cluster initialization: Initialize the cluster center and set the parameters required in the clustering process, including the objective function difference threshold, fuzzy factor and maximum number of iterations; Step S2: Calculate the decision weight matrix: Calculate the membership matrix and use the decision weight function to map the membership matrix into a new decision weight matrix; Step S3: constructing a fuzzy cluster shadow set: establishing an optimization principle, using a particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster, thereby constructing a fuzzy cluster shadow set; Step S4: Update the cluster center: construct a cluster center update formula based on the contribution of different partitions in the fuzzy cluster shadow set to the cluster center; Step S5: Iterate step S2 to step S4 until clustering ends after reaching the iteration stop condition, wherein the iteration stop condition is: comparing the difference of the objective function before and after the update and stopping when the objective function difference is less than the difference threshold or when reaching the maximum number of iterations.

2. The shadow set fuzzy clustering method based on decision weight drive according to claim 1 is characterized in that: In step S1, cluster initialization includes the following steps: Step S11: Initialize cluster centers: For a data set X={x1,x2,…,x n } Randomly select k cluster centers or use K-means to initialize k fuzzy cluster centers; Step S12: Setting the parameters required in the clustering process, including the objective function difference threshold ε, the fuzzy factor m, and the maximum number of iterations T.

3. The shadow set fuzzy clustering method based on decision weight drive according to claim 2 is characterized in that: In step S2, calculating the decision weight matrix specifically includes the following steps: Step S21: Calculate the membership matrix: calculate the membership of each sample to each cluster center; Step S22: constructing a decision weight function; Step S23: Calculate the decision weight matrix.

4. The shadow set fuzzy clustering method based on decision weight drive according to claim 3 is characterized in that: In step S22, the constructed decision weight function specifically includes: Step S221: Constructing a segmented decision weight function that conforms to the fuzzy cluster distribution , expressed as: ； in, Represents the segment decision weight function The next j-th sample x j Belongs to the i-th cluster center c i The decision weight value, u ij represents the jth sample x j Belongs to the i-th cluster center c i The membership degree is 1≤j≤n, 1≤i≤k, w n , w m and w p They are the membership intervals [0,α i ), [α i ,β i ] and (β i ,1], the decision weight function is specifically expressed as: ； ； ； Among them, (α i ,β i ) is a pair of thresholds, satisfying 0<α i <β i <1, parameters ， ， ， is the shape parameter. The parameter σ is used to determine the key point (θ3, 0.5), which is used to adjust the curvature of the decision weight function. In addition, θ1=1 and θ2=0 are set. Step S222: Construct an S-type decision weight function that conforms to the fuzzy cluster distribution , expressed as: ； in, Represents the segment decision weight function The next j-th sample x j Belongs to the i-th cluster center c i The decision weight value, u ij represents the jth sample x j Belongs to the i-th cluster center c i The membership degree is 1≤j≤n, 1≤i≤k.

5. The shadow set fuzzy clustering method based on decision weight drive according to claim 4 is characterized in that: In step S3, the specific steps of constructing the fuzzy cluster shadow set are: Step S31: Establishing optimization principles: setting three mapping sets of fuzzy sets to establish optimization principles for constructing fuzzy cluster shadow sets; Step S32: using a particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster; Step S33: construct a fuzzy cluster shadow set, each fuzzy cluster is represented as c i ={Inc(c i ),shd(c i ),Exl(c i )}.

6. The shadow set fuzzy clustering method based on decision weight drive according to claim 5 is characterized in that: In step S31, the specific steps of establishing the optimization principle are: Step S311: Determine the three-value set of fuzzy set mapping: for any fuzzy cluster center c i , the decision weight values ​​of all samples are regarded as a fuzzy set, represented by w i ={w ij |j=1,2,…,n}, construct fuzzy set w i The three mapping sets on : ； Among them, (α i ,β i ) is a pair of thresholds, satisfying 0<α i <β i <1,w ij Represents sample x j Belongs to cluster center c i The decision weight value of Step S312: Calculate the partition threshold (α i ,β i ), using the minimum error optimization principle: ; Among them, L is the target optimization function, w ij Represents sample x j Belongs to cluster center c i The decision weight value, and w ij ∈[0,1], which can be specifically expressed as and .

7. The shadow set fuzzy clustering method based on decision weight drive according to claim 6 is characterized in that: In step S4, the specific steps of updating the cluster center are: For any fuzzy cluster c i , whose blurred shadow is represented by c i ={Inc(c i ),shd(c i ),Exl(c i )}, according to the difference in contribution of samples in different areas of the fuzzy cluster shadow set to the update of the cluster center, the following cluster center update formula is constructed based on the decision weight matrix: ; Among them, card is the cardinality, X is the data set, and X={x1,x2,…,x n },(α i ,β i ) is the partition threshold of each fuzzy cluster, w ij Represents sample x j Belongs to cluster center c i The decision weight value of .

8. The shadow set fuzzy clustering method based on decision weight drive according to claim 7 is characterized in that: In step S5, the iteration stop condition is specifically: (1) Compare the difference between the objective functions J(t) and J(t-1) updated at the t-th iteration and the t-1-th iteration. If J(t)-J(t-1)≤ε; The iteration stops, Among them, ε is the objective function difference threshold; Among them, the objective function of each iteration is expressed as follows: ； Where W represents the decision weight value matrix induced by the membership matrix, C represents the cluster center set, and C=(c1,c2,…,c k ), m is the fuzzification parameter and m≥1, d ij Represents sample x j With cluster center c i Euclidean distance, 1≤j≤n, 1≤i≤k, w ij Represents sample x j Belongs to cluster center c i The decision weight value, and w ij ∈[0,1]; (2) Or the process stops when the number of iterations reaches the maximum number of iterations T.

9. A shadow set fuzzy clustering system based on decision weight drive, to implement a shadow set fuzzy clustering method based on decision weight drive as claimed in claim 1, characterized in that: include: Initialization module: used to initialize the cluster center and set the parameters required in the clustering process; Decision weight matrix calculation module: used to calculate the membership matrix and use the decision weight function to map the membership matrix into a new decision weight matrix; Constructing fuzzy cluster shadow set module: used to establish optimization principles and use particle swarm optimization algorithm to calculate the partition threshold of each fuzzy cluster, so as to construct fuzzy cluster shadow set; Cluster center update module: used to construct the cluster center update formula according to the contribution of different partitions in the fuzzy cluster shadow set to the cluster center; Check iterative stop condition module: used for iterative calculation, compares the difference of objective function before and after update and stops when the difference of objective function is less than the difference threshold or when the maximum number of iterations is reached.

Images