Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

Fuzzy Deep Neural Network for classification of overlapped data

Dabare, R., Wong, K.W., Shiratuddin, M.F. and Koutsakis, P.ORCID: 0000-0002-4168-0888 (2019) Fuzzy Deep Neural Network for classification of overlapped data. Lecture Notes in Computer Science, 11953 . pp. 633-643.

Link to Published Version:
*Subscription may be required


Deep Learning is a popular and promising technique for classification problems. This paper proposes the use of fuzzy deep learning to improve the classification capability when dealing with overlapped data. Most of the research focuses on classification and uses traditional truth and false criteria. However, in reality, a data item may belong to different classes at different degrees. Therefore, the degree of belonging of each data item to a class needs to be considered for classification purposes in some cases. When a data item belongs to different classes with different degrees, then there exists an overlap between the classes. For this reason, this paper proposes a Fuzzy Deep Neural Network based on Fuzzy C-means clustering, fuzzy membership grades and Deep Neural Networks to address the over-lapping issue focused on binary classes and multi-classes. The proposed method converts the original attribute values to relevant cluster centres using the proposed Fuzzy Deep Neural Network. It then trains them with the original output class values. Thereafter, the test data is checked with the Fuzzy Deep Neural Network model for its performance. Using three popular datasets in overlapped and fuzzy data literature, the method presented in this paper outperforms the other methods compared in this study, which are Deep Neural Networks and Fuzzy classification.

Item Type: Journal Article
Murdoch Affiliation(s): Information Technology, Mathematics and Statistics
Publisher: Springer Verlag
Copyright: © 2019 Springer Nature Switzerland AG
Item Control Page Item Control Page