A Novel Double-Index-Constrained, Multi-View, Fuzzy-Clustering Algorithm and its Application for Detecting Epilepsy Electroencephalogram Signals

Jiaqi Zhu,Yizhang Jiang,Pengjiang Qian,Xiaoqing Gu,Kaijian Xia,Shi Qiu,Jing Xue,Kang Li

doi:10.1109/access.2019.2931695

Abstract

When processing a multi-view, epilepsy electroencephalogram (EEG) dataset, the traditional single-view clustering algorithms cannot fully mine the correlation information between each view and identify the importance of each view because of the limitations of its own methods. This limitation causes poor clustering performance when using these classic, single-view clustering algorithms. To solve this problem, a novel double-index-constrained, multi-view, fuzzy clustering algorithm (DIC-MV-FCM) is proposed for the automatic detection of epilepsy EEG data. The DIC-MV-FCM algorithm is integrated into the multi-view clustering technology and the view-weighted adaptive learning strategy, which can effectively use the correlation information between each view and control the importance of each view to improve the final clustering performance. The experimental results using several epilepsy EEG datasets show that the proposed DIC-MV-FCM algorithm has better clustering performance than the traditional clustering algorithms for processing multi-view EEG data.

Highlights

Epilepsy is a common and frequent brain disease with transient and repetitive disorders of the central nervous system due to excessive discharge of brain lesions [1]
Even though the ensemble learning strategy was introduced at the end to obtain a global clustering result, the interaction between each view was not used well because the interactive learning between each view was neglected. This ensemble learning strategy will cause the final clustering performance of DI-fuzzy c-means (FCM) to suffer from the impact of a certain view with a bad clustering result
Unlike the classical multiview clustering that treats all views as important, the DIC-MV-FCM weights the degree of importance of each view by using a multi-view adaptive weighting strategy

Summary

INTRODUCTION

Epilepsy is a common and frequent brain disease with transient and repetitive disorders of the central nervous system due to excessive discharge of brain lesions [1]. Compared to traditional single-view clustering algorithms, the main contributions of the proposed DIC-MV-FCM algorithm can be summarized as follows: 1) An unsupervised multi-view clustering technology that does not use labeled data is proposed to analyze EEG data instead of the traditional supervised classification methods. After performing this single-view processing of the working principle of the clustering task of multi-view data, it is not difficult to find that this clustering technology cannot fully mine the correlation information between each view and identify the importance of each view. This ensemble learning strategy will cause the final clustering performance of DI-FCM to suffer from the impact of a certain view with a bad clustering result

MULTI-VIEW CO-FKM ALGORITHM

EXPERIMENTAL STUDY

CONCLUSION