An Unsupervised feature extraction method based on self coded neural network. Issue 4 (June 2019)

• Record Type: Journal Article
• Authors: Hao, Long; Jiang, F
• Is Part Of: Journal of physics. Volume 1213:Issue 4(2019)
• Publication Date: 2019-06
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1213/4/042067
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: The way to tap the inherent law of ECG data and feature extraction is of vital in the absence of prior knowledge of the situation. This paper presents a self coded neural network model, which is reconstructed after compression of ECG data. The compression process eliminates the redundancy of data within the original data firstly from the low dimensional with more concise representations. In order to verify the features extracted from the neural network, we select the simplest and most direct distance based K nearest neighbour classification method. By using a self compiled neural network to extract features from the data set, the classification accuracy of K nearest neighbour classification can be greatly improved. The experimental results show that the unsupervised feature extraction method based on self coded neural network can effectively extract the features of the data in practical applications.