Research On Classification Algorithm Based On Fuzzy Cognitive Maps

Fuzzy cognitive maps(FCMs)is a soft computing technology that combines the advantages of fuzzy logic and recursive neural networks.It has found widespread applications in system dynamic modeling,time series prediction,decision making,and process control.However,the application of fuzzy cognitive maps in data classification has been largely neglected due to the absence of hidden neural entities,resulting in low generality and prediction accuracy of the classification models.In addition,the traditional method of using fuzzy cognitive maps in time series classification has limitations,including difficulties in effectively extracting data features in the presence of noise and challenges in explaining the machine learningbased classification process.To address these challenges,this paper proposes integrating capsule networks into the reasoning rules of fuzzy cognitive maps to establish a more robust fuzzy cognitive maps classification model.The paper also explores the application of fuzzy cognitive maps in bearing fault diagnosis and time series classification.The main contributions of this paper are as follows:Firstly,the paper proposes a general and simple model for classification problems based on fuzzy cognitive maps,which integrates capsule networks into the inference rules of fuzzy cognitive maps to form a new inference rule with a strong coupling coefficient.The proposed method uses particle swarm optimization algorithm to learn the weight and cross entropy with constraints as the loss function.Experimental results show that the proposed method outperforms existing methods under the same framework.Secondly,the paper proposes a two-stage fuzzy cognitive maps classification method for time series classification,which can accurately and quickly classify time series data.In the feature extraction stage,singular spectrum analysis is used to separate noise and extract effective information from time series.At the same time,univariate time series can be converted into multivariable time series.Then,the fuzzy cognitive maps is used to model the time series,which can extract features quickly and effectively.In the classification stage,a fuzzy cognitive maps based classification model is established to classify the feature matrix.The proposed method is called the two-stage fuzzy cognitive maps.Experimental results show that this method can extract features from noisy data more effectively,and outperforms existing methods.Furthermore,the system modeling process is transparent and interpretable.Lastly,the paper presents a method for bearing fault diagnosis by modeling the problem as a time series classification problem.The method extends one-dimensional time series to multidimensional time series using singular spectrum analysis to reduce noise.The convex optimization algorithm is then used to realize fast and robust learning of the FCMs model.Finally,the feature matrix is classified using the FCMs classifier,and the proposed method is validated on the Case Western Reserve University(CWRU)bearing data set.Experimental results demonstrate the superior performance of the proposed method on the bearing fault data set,providing a new reference for the field of bearing fault diagnosis.In summary,this paper proposes several novel methods for fuzzy cognitive mapbased classification and time series classification,as well as a new approach for bearing fault diagnosis.The proposed methods are shown to outperform existing methods and provide a transparent and interpretable system modeling process.The contributions of this paper will be valuable for researchers and practitioners in the field of soft computing and data analysis.