Data classification method and system based on improved TSK fuzzy classification model

Abstract

The invention relates to the technical field of data classification, and discloses a data classification method and system based on an improved TSK fuzzy classification model, and the method comprises the steps: obtaining a to-be-classified data set, and dividing the to-be-classified data set into a training set and a test set; based on the training set, utilizing a supervised enhanced soft subspace clustering algorithm and a deep learning algorithm to respectively optimize a fuzzy rule consequent and a fuzzy rule consequent of a fuzzy rule of a pre-constructed TSK fuzzy classification model; and inputting the test set into the optimized TSK fuzzy classifier model for classification processing to obtain a data classification result. According to the method, the fuzzy rule consequent and fuzzy rule consequent parameters of the TSK fuzzy classification model are optimized respectively, so that the adaptability of the model to high-dimensional data and complex data sets is improved, and the precision and efficiency of the model in data classification are further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image classification, and more specifically, to a data classification method and system based on an improved TSK fuzzy classification model. Background Art

[0002] With the development of artificial intelligence and big data technology, fuzzy systems have been widely used in tasks such as classification and regression. Among them, the Takagi-Sugeno-Kang (TSK) fuzzy system has become an important classifier due to its excellent interpretability and ability to model data uncertainty. The TSK fuzzy system implements an uncertain reasoning process similar to human reasoning through an IF-THEN model based on fuzzy rules, and performs well in classification and regression problems. Compared with other fuzzy systems, the TSK fuzzy system outputs a certain value, making it easier to interpret and understand, so it is widely used in practical applications. To further improve the performance of the model, existing technologies combine a variety of optimization methods, such as optimizing fuzzy rule antecedents through fuzzy C-means clustering (FCM) and mean square hierarchical fuzzy clustering (SFCM), and optimizing fuzzy rule consequents through genetic algorithms (GA), deep convolution (DC), and support vector machines (MSVM). In addition, deep neural networks and new loss functions have been introduced in recent years to improve fuzzy rule generation and generalization performance.

[0003] However, in practical applications, the existing TSK fuzzy classification model has high-dimensional features that make the fuzzy rules lengthy and complex, which weakens the interpretability and accuracy of the model; too many fuzzy rules easily introduce redundancy and reduce generalization; in addition, the output formula based on the softmax function has saturation, which leads to an imbalance in the activation level of fuzzy rules, and only a few fuzzy rules are frequently used, which increases the complexity of the model. The above problems all lead to the defects of low classification accuracy and low efficiency in data classification of the TSK fuzzy classification model. Summary of the invention

[0004] In order to overcome the defects of low classification accuracy and low efficiency in data classification of the existing TSK fuzzy classification model, the present invention proposes the following technical solutions: In a first aspect, the present invention proposes a data classification method based on an improved TSK fuzzy classification model, comprising: Obtain a data set to be classified, and divide the data set to be classified into a training set and a test set; Based on the training set, using a supervised enhanced soft subspace clustering algorithm and a deep learning algorithm, respectively optimizing the fuzzy rule antecedent and the fuzzy rule consequent of the fuzzy rules of the pre-constructed TSK fuzzy classification model; The test set is input into the optimized TSK fuzzy classifier model for classification processing to obtain the data classification result.

[0005] As a preferred technical solution, before inputting the test set into the optimized TSK fuzzy classifier model for classification processing, the method further includes: The improved fuzzification operation is used to homogenize the triggering level of the fuzzy rule, and its expression is as follows:

[0006] in, Fuzzy rules r The trigger value of r The trigger level on the current input sample; is the dimension of the input feature, Indicates that the input sample is d The value of the feature dimension, Representing fuzzy rules r In the d The cluster centers on the feature dimensions, Representing fuzzy rules r In the d The variance on the feature dimension, Represents the maximum value among all fuzzy rule trigger values.

[0007] As a preferred technical solution, the pre-built TSK fuzzy classification model includes an input layer, a fuzzification layer, a fuzzy rule layer, an inference layer and an output layer; The input samples are fed into the fuzzification layer through the input layer; The fuzzification layer calculates the membership value based on the feature value of the input sample, the cluster center and feature range of the fuzzy rule; The fuzzy rule layer triggers the corresponding fuzzy rules according to the membership value output by the fuzzification layer; The reasoning layer performs fuzzy reasoning based on the triggered fuzzy rules to obtain the normalized trigger value of the fuzzy rules; The output layer performs defuzzification according to the normalized trigger value of the fuzzy rule to obtain the final classification result.

[0008] As a preferred technical solution, the reasoning layer performs fuzzy reasoning according to the triggered fuzzy rules to obtain the normalized trigger value of the fuzzy rules. , whose expression is as follows:

[0009] Where R is the total number of fuzzy rules, Represents fuzzy rules r Trigger value Perform exponential operation.

[0010] As a preferred technical solution, the output layer performs defuzzification processing according to the normalized trigger value of the fuzzy rule to obtain the final classification result, which is expressed as follows:

[0011] in, represents the predicted value of the input sample x in category c, Represents the output value of fuzzy rule r on category c.

[0012] As a preferred technical solution, the supervised enhanced soft subspace clustering algorithm is used to optimize the fuzzy rule antecedents of the fuzzy rules of the pre-built TSK fuzzy classification model, including: Initialize the fuzzy membership matrix of the fuzzy rule antecedents , cluster center matrix , weight matrix and the probability matrix .

[0013] For training set data samples and its corresponding sample labels , construct the objective function of the supervised enhanced soft subspace clustering algorithm , whose expression is as follows:

[0014] in, N is the number of training set data samples, R is the number of cluster centers, D is the data feature dimension, C is the number of classification categories, Indicates n Sample pair r The membership degree of the cluster center, m>1 is the fuzzy index, Representation sample Belongs to category c The label value of Represents the cluster center r In the feature dimension d The weight on Represents the cluster center r In category c The probability value on For the n input samples, For the r Cluster centers, is the cluster center of all data samples in the training set, For the r The weight of the cluster center, is the adjustment item parameter, Iteratively optimize the fuzzy membership matrix , cluster center matrix , weight matrix and the probability matrix , until the objective function The convergence conditions are met and the optimized fuzzy rule antecedents are obtained.

[0015] As a preferred technical solution, the fuzzy membership matrix is ​​iteratively optimized by the Lagrange multiplier method. , cluster center matrix , weight matrix and the probability matrix , until the objective function The convergence condition is satisfied and its expression is as follows:

[0016]

[0017]

[0018] .

[0019] As a preferred technical solution, a deep learning algorithm is used to optimize the fuzzy rule consequences of the fuzzy rules of the pre-built TSK fuzzy classification model, including: Initialize the linear weight vector and bias term of the fuzzy rule consequent; Perform batch normalization on the training set sample data, and apply the Sigmoid function to the normalized data for nonlinear activation; Using the training set sample data that has been normalized and nonlinearly activated, the parameters of the fuzzy rule consequent are trained and optimized until the optimization objective function of the fuzzy rule consequent meets the convergence condition; Among them, the expression of the optimization objective function of the fuzzy rule consequent is as follows:

[0020] in, Representation sample n Belongs to category c The true label of Representation sample Predicted as category c The probability of represents the normalized activation weight of the fuzzy rule r, represents the linear weight vector of fuzzy rule r for category c, represents a linear weight vector The transpose of For sample The nonlinear embedding result represents the sample Through batch normalization Processing and activation functions The subsequent feature map.

[0021] As a preferred technical solution, in the training and optimization of the parameters of the fuzzy rule consequent, the method further includes: Assign a retention probability to each fuzzy rule; In each training optimization iteration, a random number in the interval [0, 1] is generated for each fuzzy rule; If the generated random number is less than or equal to the retention probability set for the fuzzy rule, the fuzzy rule is retained in the current iteration, otherwise, the fuzzy rule is discarded; Only the fuzzy rules that are not discarded are trained and their parameters updated.

[0022] In a second aspect, the present invention further proposes a data classification system based on an improved TSK fuzzy classification model, which is applied to the data classification method based on the improved TSK fuzzy classification model as described in any solution of the first aspect, comprising: An acquisition and division module is used to acquire a data set to be classified and divide the data set to be classified into a training set and a test set; An optimization module, for respectively optimizing the fuzzy rule antecedent and the fuzzy rule consequent of the fuzzy rules of the pre-built TSK fuzzy classification model based on the training set by using a supervised enhanced soft subspace clustering algorithm and a deep learning algorithm; The classification module is used to input the test set into the optimized TSK fuzzy classifier model for classification processing to obtain the data classification results.

[0023] The beneficial effects of the present invention include at least: The present invention optimizes the parameters of the fuzzy rule antecedent and fuzzy rule consequent of the TSK fuzzy classification model by using the supervised enhanced soft subspace clustering algorithm and the deep learning algorithm, thereby improving the model's adaptability to high-dimensional data and complex data sets. The supervised enhanced soft subspace clustering algorithm can select different feature subspaces for different fuzzy rules, reduce the influence of irrelevant or redundant features, and thus improve the accuracy and classification accuracy of fuzzy rules; the deep learning algorithm further improves the expression ability of the fuzzy rule consequent through effective parameter optimization methods (such as batch gradient descent), ensuring that the model has better generalization ability and stability when processing large-scale data sets. These optimizations work together to enable the model to achieve higher classification accuracy with fewer computing resources in data classification tasks, while significantly improving classification efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flowchart of a data classification method based on an improved TSK fuzzy classification model provided in this embodiment.

[0025] Figure 2 A comparison chart of average normalized trigger level results on the entire Clave dataset using the MTSK fuzzification operation and not using the MTSK fuzzification operation provided by an embodiment of the present invention.

[0026] Figure 3 A comparison diagram of average normalized trigger level results on the entire data set using MTSK fuzzification operation on the WDBC data set and not using MTSK fuzzification operation provided by an embodiment of the present invention.

[0027] Figure 4 A schematic diagram of a fuzzy rule antecedent framework provided in an embodiment of the present invention.

[0028] Figure 5 This is a diagram showing the effect of reducing the number of regular features of the SESSC algorithm provided by an embodiment of the present invention on a DRD data set.

[0029] Figure 6 This is a comparison chart of the number of model parameters and RCA values ​​obtained when SESSC is not used and when SESSC is used provided in an embodiment of the present invention.

[0030] Figure 7 A schematic diagram of the principle of discarding fuzzy rules using the dropout technology provided in an embodiment of the present invention.

[0031] Figure 8 A comparison chart of the number of model parameters and RCA values ​​obtained when the deep learning optimization technology is not used and when the deep learning optimization technology is used provided in an embodiment of the present invention.

[0032] Fig. 9 A schematic diagram of a partial structure of a SESSC-DL-MTSK model provided in an embodiment of the present invention.

[0033] Fig.10 This is a diagram showing the effect of reducing the number of regular features on a DRD data set without using deep learning optimization technology provided by an embodiment of the present invention.

[0034] Fig.11 A diagram showing the effect of reducing the number of regular features on a DRD data set using deep learning optimization technology provided in an embodiment of the present invention.

[0035] Fig.12 This is an architecture diagram of a data classification system based on an improved TSK fuzzy classification model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred technical solutions. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred technical solutions are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0037] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0038] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0039] In order to facilitate understanding of the embodiments of the present invention, before describing each embodiment of the present invention, some concepts involved in all the embodiments of the present invention are first appropriately explained.

[0040] TSK fuzzy system (TSKFS) is a classic fuzzy reasoning model. For TSKFS, the definition of fuzzy rules is as follows:

[0041] In the formula, is the D-dimensional input vector in the training dataset, It is r The fuzzy rule d The fuzzy subsets related to the features are and It is c The consequent parameters of fuzzy rules of categories.

[0042] The fuzzy algorithm of the present invention can use different membership functions, among which the Gaussian membership function has a smooth shape, which makes it less sensitive to small changes in input data and helps to reduce the sensitivity of the output of the fuzzy system to input noise. Therefore, the present invention adopts the Gaussian membership function, which has the following characteristics: In the fuzzy set The membership value on is:

[0043] In the formula, the parameters and It can be calculated by clustering algorithm. Use FCM (Fuzzy C-means, fuzzy C-means clustering algorithm) to calculate and , can be obtained by the following formula:

[0044]

[0045] In the formula, N is the number of input data, Represents the input data in the rth cluster j The membership value of . h is a manually adjustable scale parameter.

[0046] The output of the TSK fuzzy classifier can be expressed by the following formula:

[0047] in,

[0048]

[0049] is the trigger level of the rth fuzzy rule. In order to optimize the TSK fuzzy classifier, it is necessary to fine-tune the parameters of the antecedent and consequent of the fuzzy rules.

[0050] Example 1 This embodiment proposes a data classification method based on an improved TSK fuzzy classification model. Figure 1 As shown, Figure 1 A flow chart of a data classification method based on an improved TSK fuzzy classification model provided in this embodiment, the method comprises the following steps: S1: Obtain a data set to be classified, and divide the data set to be classified into a training set and a test set.

[0051] S2: Based on the training set, the fuzzy rule antecedents and fuzzy rule consequents of the fuzzy rules of the pre-built TSK fuzzy classification model are optimized using a supervised enhanced soft subspace clustering algorithm and a deep learning algorithm.

[0052] S3: Input the test set into the optimized TSK fuzzy classifier model for classification processing to obtain the data classification result.

[0053] It can be understood that the present invention optimizes the fuzzy rule antecedent and fuzzy rule consequent parameters of the TSK fuzzy classification model by using the supervised enhanced soft subspace clustering algorithm and the deep learning algorithm, thereby improving the model's adaptability to high-dimensional data and complex data sets. The supervised enhanced soft subspace clustering algorithm can select different feature subspaces for different fuzzy rules, reduce the impact of irrelevant or redundant features, and thus improve the accuracy and classification accuracy of fuzzy rules; the deep learning algorithm further improves the expression ability of the fuzzy rule consequent through effective parameter optimization methods (such as batch gradient descent), ensuring that the model has better generalization ability and stability when processing large-scale data sets. These optimizations work together to enable the model to achieve higher classification accuracy with fewer computing resources in data classification tasks, while significantly improving classification efficiency.

[0054] Example 2 This embodiment makes improvements on the data classification method based on the improved TSK fuzzy classification model proposed in Embodiment 1.

[0055] In this embodiment, the triggering level of the fuzzy rule is uniformly processed by using an improved fuzzification operation, which specifically includes: make for ,but can be rewritten as:

[0056] The final defuzzification operation can be calculated using the softmax function. However, the softmax function is saturated, so the defuzzification operation will also be saturated. This means that as the dimension increases, The smaller it is, the fewer fuzzy rules each input can trigger, which leads to saturation of defuzzification. For example, suppose the model has four fuzzy rules, ,So , at this time, the activation level difference of each fuzzy rule is very small, which means that more fuzzy rules can be triggered; when the dimension increases, , , at this time, only the first fuzzy rule can be triggered, and the rest of the fuzzy rules are hardly triggered.

[0057] In order to avoid the saturation problem of fuzzy rules, the present invention proposes an improved fuzzification operation, named MTSK, which is expressed as follows:

[0058] in, Fuzzy rules r The trigger value ofr The trigger level on the current input sample; is the dimension of the input feature, Indicates that the input sample is d The value of the feature dimension, Representing fuzzy rules r In the d The cluster centers on the feature dimensions, Representing fuzzy rules r In the d The variance on the feature dimension, Represents the maximum value among all fuzzy rule trigger values.

[0059] In this embodiment, the pre-built TSK fuzzy classification model includes an input layer, a fuzzification layer, a fuzzy rule layer, an inference layer and an output layer; The input samples are fed into the fuzzification layer through the input layer; The fuzzification layer calculates the membership value based on the feature value of the input sample, the cluster center and feature range of the fuzzy rule; The fuzzy rule layer triggers the corresponding fuzzy rules according to the membership value output by the fuzzification layer; The reasoning layer performs fuzzy reasoning based on the triggered fuzzy rules to obtain the normalized trigger value of the fuzzy rules; The output layer performs defuzzification according to the normalized trigger value of the fuzzy rule to obtain the final classification result.

[0060] In this embodiment, the reasoning layer performs fuzzy reasoning according to the triggered fuzzy rules to obtain the normalized trigger value of the fuzzy rules. , whose expression is as follows:

[0061] Where R is the total number of fuzzy rules, Represents fuzzy rules r Trigger value Perform exponential operation.

[0062] In this embodiment, the output layer performs defuzzification processing according to the normalized trigger value of the fuzzy rule to obtain the final classification result, which is expressed as follows:

[0063] in, Represents the predicted value of the input sample x in category c, Represents the output value of fuzzy rule r on category c.

[0064] As an exemplary illustration, in order to prove that the MTSK fuzzification operation proposed in the present invention can alleviate the fuzzy rule saturation problem, Figure 2 and Figure 3 As shown, Figure 2 A comparison chart of the average normalized trigger level results of using MTSK fuzzification operation on the Clave dataset and not using MTSK fuzzification operation on the entire dataset provided by an embodiment of the present invention, Figure 3 The comparison diagram of the average normalized trigger level results of using MTSK fuzzification operation on the WDBC data set and not using MTSK fuzzification operation on the entire data set provided by the embodiment of the present invention is shown in FIG. Figure 2 and Figure 3 It can be seen that for the model without MTSK, the saturated fuzzy rules have a larger average triggering level than other fuzzy rules, so the contribution of fuzzy rules to the output is significantly different. The average normalized fuzzy rule excitation level of the model using MTSK is more uniform than the average normalized fuzzy rule excitation level of the model without MTSK, which also leads to better classification performance.

[0065] In this embodiment, the fuzzy rule antecedents of the fuzzy rules of the pre-constructed TSK fuzzy classification model are optimized using the supervised enhanced soft subspace clustering algorithm, including: Initialize the fuzzy membership matrix of the fuzzy rule antecedents , cluster center matrix , weight matrix and the probability matrix ; For training set data samples and its corresponding sample labels , construct the objective function of the supervised enhanced soft subspace clustering algorithm , whose expression is as follows:

[0066] in, N is the number of training set data samples, R is the number of cluster centers, D is the data feature dimension, C is the number of classification categories, Indicates n Sample pair r The membership degree of the cluster center, m>1 is the fuzzy index, Representation sample Belongs to category c The label value of Represents the cluster center r In the feature dimension d The weight on Represents the cluster center r In category c The probability value on For the n input samples, For the r Cluster centers, is the cluster center of all data samples in the training set, For the r The weight of the cluster center, is the adjustment item parameter, Iteratively optimize the fuzzy membership matrix , cluster center matrix , weight matrix and the probability matrix , until the objective function The convergence conditions are met and the optimized fuzzy rule antecedents are obtained.

[0067] In this embodiment, the fuzzy membership matrix is ​​iteratively optimized by the Lagrange multiplier method. , cluster center matrix , weight matrix and the probability matrix , until the objective function The convergence condition is satisfied and its expression is as follows:

[0068]

[0069]

[0070] .

[0071] As an exemplary explanation, because the core of the TSK fuzzy system is to build a fuzzy rule base, each fuzzy rule is divided into a fuzzy rule antecedent and a fuzzy rule consequent, so the fuzzy rule antecedent part of the fuzzy rule must be constructed first. However, because most TSK fuzzy systems currently use FCM clustering algorithms or new feature-weighted FCM clustering algorithms, entropy-weighted clustering algorithms, etc., they cannot handle multi-dimensional feature spaces well and cannot propose potential feature connections. The present invention uses the supervised enhanced soft subspace clustering algorithm SESSC to process the features of the input data and train effective fuzzy rule antecedents. This embodiment uses the pytorch library in python to construct a fuzzy rule antecedent framework, such as Figure 4 As shown, Figure 4 The principle diagram of constructing a fuzzy rule antecedent framework provided by an embodiment of the present invention specifically includes: Calculate the sample center of all input samples , use the ordinary FCM clustering algorithm to calculate the original cluster center matrix V, initialized according to the dimension , initialized according to the number of categories of data samples .

[0072] Use pytorch to write the SESSC function, and implement the fuzzy membership matrix inside the SESSC function , cluster center matrix , weight matrix and the probability matrix The iterative code of the update formula returns a trained cluster center matrix after continuous iterative updates and reaching the threshold. V .

[0073] This embodiment builds a fuzzy rule library containing 20 fuzzy rules, and each fuzzy rule has a different cluster center matrix. Take the medical data DRD in the medical field as an example to check the number of data sample features after using the SESSC algorithm. The DRD data set has 1151 data samples, the number of features is 19, and the number of categories is 2. Use the matplotlib library to visualize the cluster center matrix after the SESSC algorithm is processed to see whether the number of features used by each fuzzy rule has been successfully streamlined, such as Figure 5 As shown, Figure 5 The effect diagram of reducing the number of rule features of the SESSC algorithm provided by the embodiment of the present invention on the DRD data set shows that when the original number of features is 19, the number of features used in the 20 fuzzy rules is less than twice the original number, and the maximum number of features used is only 9. This shows that each fuzzy rule has achieved good rule simplification, can well mine the potential relationship between features, remove redundant features, and finally form a streamlined and efficient fuzzy rule base.

[0074] In order to prove the effectiveness of SESSC, by keeping other parameters unchanged, the complexity and classification performance of the model are analyzed when SESSC is used and not used. Figure 6 As shown, Figure 6 A comparison chart of the number of model parameters and RCA values ​​obtained when SESSC is not used and when SESSC is used provided in an embodiment of the present invention. Figure 6It can be seen that after using SESSC, the model complexity of each data set has been reduced to varying degrees, indicating that the use of features is different. And when the data dimension is large, such as the data sets Satellite and Usps, the model accuracy is higher than that without SESSC. Although the RCA values ​​of the data sets DRD and Steel have dropped significantly, it is because too few features are used. From this, it can be concluded that SESSC can effectively reduce the number of parameters and has better classification performance when processing high-dimensional data, helping to build a simple and efficient fuzzy system.

[0075] In this embodiment, the fuzzy rule consequence of the fuzzy rule of the pre-built TSK fuzzy classification model is optimized by using a deep learning algorithm, including: Initialize the linear weight vector and bias term of the fuzzy rule consequent; Perform batch normalization on the training set sample data, and apply the Sigmoid function to the normalized data for nonlinear activation; Using the training set sample data that has been normalized and nonlinearly activated, the parameters of the fuzzy rule consequent are trained and optimized until the optimization objective function of the fuzzy rule consequent meets the convergence condition; Among them, the expression of the optimization objective function of the fuzzy rule consequent is as follows:

[0076] in, Representation sample n Belongs to category c The true label of Representation sample Predicted as category c The probability of represents the normalized activation weight of the fuzzy rule r, represents the linear weight vector of fuzzy rule r for category c, represents a linear weight vector The transpose of For sample The nonlinear embedding result represents the sample Through batch normalization Processing and activation functions The subsequent feature map.

[0077] Using BN to normalize the input data can speed up the model training by reducing internal covariates and alleviate the gradient vanishing problem. , the output of BN is:

[0078] in:

[0079]

[0080] In the formula, and are the mean and standard deviation of small batch samples, and It is a parameter that can be learned during the training process. If the data after standardization is not suitable, it can be slightly offset to make it suitable. It is usually set to 1e-8 to avoid the situation where the denominator is 0. After BN processing the original input data, the obtained Nonlinear processing is performed to enhance the ability to detect and characterize the nonlinear structures and features hidden in the data. The present invention uses the sigmoid function as the nonlinear function, and the data after nonlinear changes is , whose expression is as follows:

[0081] Therefore, the output of the TSK fuzzy classifier can also be rewritten as:

[0082] In this embodiment, in the training and optimization of the parameters of the fuzzy rule consequent: Assign a retention probability to each fuzzy rule; in each training optimization iteration, generate a random number in the interval [0, 1] for each fuzzy rule; if the generated random number is less than or equal to the retention probability set for the fuzzy rule, then retain the fuzzy rule in the current iteration, otherwise, discard the fuzzy rule; only train and update the parameters of the fuzzy rules that are not discarded. Figure 7 As shown, Figure 7 The schematic diagram of using the dropout technology to discard fuzzy rules provided by the embodiment of the present invention is as follows: Figure 7 The second fuzzy rule is discarded.

[0083] As an exemplary explanation, in order to prove the effectiveness of deep learning optimization technology, this embodiment analyzes the complexity and classification performance of the model when deep learning optimization technology is not used and when deep learning optimization technology is used by keeping other parameters unchanged. Figure 8 As shown, Figure 8 A comparison chart of the number of model parameters and RCA values ​​obtained when the deep learning optimization technology is not used and when the deep learning optimization technology is used provided in the embodiment of the present invention. Figure 8It can be seen that the model that removes redundant rules has a reduced number of parameters, which improves the generalization of the model and greatly improves the classification performance of the model overall.

[0084] Finally, the schematic diagram of the partial structure of the improved TSK fuzzy classification model obtained in this embodiment is as follows: Fig. 9 As shown in the figure, named SESSC-DL-MTSK model, the figure shows that the antecedent of the fuzzy rule is generated and optimized using SESSC (supervised enhanced soft subspace clustering algorithm), the consequent can be generated and optimized using DL (deep learning technology), and the fuzzification operation is performed using MTSK.

[0085] Example 3 This embodiment, based on the data classification method based on the improved TSK fuzzy classification model proposed in Example 2, performs classification tests on medical data, food data, satellite data and interpersonal communication data to verify the effectiveness of the SESSC-DL-MTSK model proposed in the present invention.

[0086] In this embodiment, the proposed SESSC-DL-MTSK model is evaluated from the two perspectives of classification performance and model complexity. It can have good classification performance and Robustness when processing high-dimensional data and large-scale data, and ensure that the activation level of each fuzzy rule in the fuzzy rule base constructed by the model is relatively average, and there will be no situation where only part of the rules are activated. While ensuring good performance, the entire model is less complex and simpler than the prior art model.

[0087] In this example, 10 classification data sets from different fields were selected from the UCI machine learning database Repository and the NTU machine learning database Lib to evaluate the proposed model algorithm. For each data set, 70% of the samples were randomly selected as the training set, and the remaining 30% of the samples were used as the test set. Table 1 gives the detailed information of the actual data set, including the name of the data set, the number of samples and features in the data set, and the number of categories.

[0088] Table 1 Actual data set information

[0089] This embodiment uses three indicators: original classification accuracy (RCA), balanced classification accuracy (BCA), and Jaccard-measure, and uses five-fold cross validation to evaluate classification performance. RCA is the total number of correctly classified test samples divided by the total number of test samples, and BCA is the average of RCA for each category, which is used to evaluate the performance of the model on an unbalanced data set. Jaccard-measure is used to evaluate the Robustness of processing an unbalanced data set.

[0090] Table 2 Comparison of average RCA of six algorithms on 10 datasets

[0091] Table 3 Average BCA comparison of six algorithms on 10 datasets

[0092] Table 4 Comparison of average Jaccard index of six algorithms on 10 datasets

[0093] The average test RCA, BCA and Jaccard-measure are shown in Tables 2, 3 and 4 respectively, and the maximum value (best performance) on each dataset is in bold. As can be seen from Table 2, the RCA index of the proposed method on the 7 datasets is better than that of other algorithms, especially for datasets with dimensions greater than 30. Further, Tables 3 and 4 give the results of BCA and Jaccard-measure, which shows that the proposed method also shows superiority and better Russin on imbalanced datasets. On average, the proposed algorithm is better than the FCM-based algorithms (TSK-PCA, TSK-UR-BN, LogTSK) in processing high-dimensional data.

[0094] The SESSC-DL-MTSK of this embodiment has lower model complexity. The experiment verifies this advantage from two aspects: quantitative evaluation of model complexity and intuitive display based on fuzzy rule form.

[0095] The model complexity of SESSC-DL-MTSK in this embodiment is calculated based on the model with the best BCA value in Table 3, and the results are shown in Table 5. It can be concluded from the results that the invented model is superior to other existing models on the 6 data sets. More importantly, the model complexity of the proposed method is significantly reduced, and the average number of parameters on the 10 data sets is the smallest.

[0096] Table 5 Comparison of model complexity of six algorithms on 10 data sets

[0097] Take the DRD dataset as an example. Fig.10 and Fig.11 As shown, Fig.10 This is a diagram showing the effect of reducing the number of regular features on a DRD data set without using deep learning optimization technology provided by an embodiment of the present invention. Fig.11The effect diagram of reducing the number of rule features on the DRD data set using deep learning optimization technology provided by the embodiment of the present invention. The x-axis represents the index of all fuzzy rules in the fuzzy rule base, and there are 20 fuzzy rules at the beginning. The y-axis represents the length of the fuzzy rule, which is determined by the number of features selected by each fuzzy rule. Fig. 9 and Fig.10 It can be seen that only some features are used after applying SESSC, that is, the length of fuzzy rules is reduced. Considering that human cognition is limited, the proposed method produces more interpretable results. In addition, after applying the deep learning optimization algorithm, some redundant fuzzy rules are replaced. Fig.10 The indices of the replaced fuzzy rules are 5, 6, 7, 9, 12 and 16 respectively, while for Fig.11 The indices of the replaced fuzzy rules are 5, 6, 7, 9, 12, 13, 16 and 18. Therefore, the SESSC-DL-MTSK proposed in the present invention can construct a fuzzy system with lower model complexity and improve interpretability.

[0098] Example 4 like Fig.12 As shown, this embodiment proposes a data classification system based on an improved TSK fuzzy classification model, which is applied to the data classification method based on the improved TSK fuzzy classification model as described in the above embodiment, including: a division module 100, an optimization module 200 and a classification module 300.

[0099] Among them, the acquisition and division module 100 is used to obtain the data set to be classified, and divide the data set to be classified into a training set and a test set; the optimization module 200 is used to optimize the fuzzy rule antecedents and fuzzy rule consequents of the fuzzy rules of the pre-constructed TSK fuzzy classification model based on the training set by using the supervised enhanced soft subspace clustering algorithm and the deep learning algorithm; the classification module 300 is used to input the test set into the optimized TSK fuzzy classifier model for classification processing to obtain the data classification result.

[0100] It should be noted that the aforementioned explanation of the data classification method embodiment based on the improved TSK fuzzy classification model is also applicable to the data classification system based on the improved TSK fuzzy classification model of this embodiment, and will not be repeated here.

[0101] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0102] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0103] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0104] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array, a field programmable gate array, etc.

[0105] A person skilled in the art may understand that all or part of the steps in the above-mentioned embodiment method may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0106] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A data classification method based on an improved TSK fuzzy classification model, characterized in that: include: Obtain a data set to be classified, and divide the data set to be classified into a training set and a test set; Based on the training set, using a supervised enhanced soft subspace clustering algorithm and a deep learning algorithm, respectively optimizing the fuzzy rule antecedent and the fuzzy rule consequent of the fuzzy rules of the pre-constructed TSK fuzzy classification model; The test set is input into the optimized TSK fuzzy classifier model for classification processing to obtain the data classification result.

2. The data classification method based on the improved TSK fuzzy classification model according to claim 1 is characterized in that: Before inputting the test set into the optimized TSK fuzzy classifier model for classification processing, the method further includes: The improved fuzzification operation is used to homogenize the triggering level of the fuzzy rule, and its expression is as follows: in, Fuzzy rules r The trigger value of r The trigger level on the current input sample; is the dimension of the input feature, Indicates that the input sample is d The value of the feature dimension, Representing fuzzy rules r In the d The cluster centers on the feature dimensions, Representing fuzzy rules r In the d The variance on the feature dimension, Represents the maximum value among all fuzzy rule trigger values.

3. The data classification method based on the improved TSK fuzzy classification model according to claim 2 is characterized in that: The pre-built TSK fuzzy classification model includes an input layer, a fuzzification layer, a fuzzy rule layer, an inference layer and an output layer; The input samples are fed into the fuzzification layer through the input layer; The fuzzification layer calculates the membership value based on the feature value of the input sample, the cluster center and feature range of the fuzzy rule; The fuzzy rule layer triggers the corresponding fuzzy rules according to the membership value output by the fuzzification layer; The reasoning layer performs fuzzy reasoning based on the triggered fuzzy rules to obtain the normalized trigger value of the fuzzy rules; The output layer performs defuzzification according to the normalized trigger value of the fuzzy rule to obtain the final classification result.

4. The data classification method based on the improved TSK fuzzy classification model according to claim 3 is characterized in that: The reasoning layer performs fuzzy reasoning based on the triggered fuzzy rules to obtain the normalized trigger value of the fuzzy rules. , whose expression is as follows: Where R is the total number of fuzzy rules, Represents fuzzy rules r Trigger value Perform exponential operation.

5. The data classification method based on the improved TSK fuzzy classification model according to claim 4 is characterized in that: The output layer performs defuzzification according to the normalized trigger value of the fuzzy rule to obtain the final classification result, which is expressed as follows: in, represents the predicted value of the input sample x in category c, Represents the output value of fuzzy rule r on category c.

6. The data classification method based on the improved TSK fuzzy classification model according to claim 1 is characterized in that: The fuzzy rule antecedents of the fuzzy rules of the pre-built TSK fuzzy classification model are optimized using the supervised enhanced soft subspace clustering algorithm, including: Initialize the fuzzy membership matrix of the fuzzy rule antecedents , cluster center matrix , weight matrix and the probability matrix ; For training set data samples and its corresponding sample labels , construct the objective function of the supervised enhanced soft subspace clustering algorithm , whose expression is as follows: in, N is the number of training set data samples, R is the number of cluster centers, D is the data feature dimension, C is the number of classification categories, Indicates n Sample pair r The membership degree of the cluster center, m>1 is the fuzzy index, Representation sample Belongs to category c The label value of Represents the cluster center r In the feature dimension d The weight on Represents the cluster center r In category c The probability value on For the n input samples, For the r Cluster centers, is the cluster center of all data samples in the training set, For the r The weight of the cluster center, is the adjustment item parameter, Iteratively optimize the fuzzy membership matrix , cluster center matrix , weight matrix and the probability matrix , until the objective function The convergence conditions are met and the optimized fuzzy rule antecedents are obtained.

7. The data classification method based on the improved TSK fuzzy classification model according to claim 6 is characterized in that: By using the Lagrange multiplier method, the fuzzy membership matrix is ​​iteratively optimized. , cluster center matrix , weight matrix and the probability matrix , until the objective function The convergence condition is satisfied and its expression is as follows: 。 8. The data classification method based on the improved TSK fuzzy classification model according to claim 1 is characterized in that: Utilize deep learning algorithms to optimize the fuzzy rule consequences of the pre-built TSK fuzzy classification model, including: Initialize the linear weight vector and bias term of the fuzzy rule consequent; Perform batch normalization on the training set sample data, and apply the Sigmoid function to the normalized data for nonlinear activation; Using the training set sample data that has been normalized and nonlinearly activated, the parameters of the fuzzy rule consequent are trained and optimized until the optimization objective function of the fuzzy rule consequent meets the convergence condition; Among them, the expression of the optimization objective function of the fuzzy rule consequent is as follows: in, Representation sample n Belongs to category c The true label of Representation sample Predicted as category c The probability of represents the normalized activation weight of the fuzzy rule r, represents the linear weight vector of fuzzy rule r for category c, represents a linear weight vector The transpose of For sample The nonlinear embedding result represents the sample Through batch normalization Processing and activation functions The subsequent feature map.

9. The data classification method based on the improved TSK fuzzy classification model according to claim 4 is characterized in that: In training and optimizing the parameters of the fuzzy rule consequent, the method further comprises: Assign a retention probability to each fuzzy rule; In each training optimization iteration, a random number in the interval [0, 1] is generated for each fuzzy rule; If the generated random number is less than or equal to the retention probability set for the fuzzy rule, the fuzzy rule is retained in the current iteration, otherwise, the fuzzy rule is discarded; Only the fuzzy rules that are not discarded are trained and their parameters updated.

10. A data classification system based on an improved TSK fuzzy classification model, characterized in that: include: An acquisition and division module is used to acquire a data set to be classified and divide the data set to be classified into a training set and a test set; An optimization module, for respectively optimizing the fuzzy rule antecedent and the fuzzy rule consequent of the fuzzy rules of the pre-built TSK fuzzy classification model based on the training set by using a supervised enhanced soft subspace clustering algorithm and a deep learning algorithm; The classification module is used to input the test set into the optimized TSK fuzzy classifier model for classification processing to obtain the data classification results.

Images