Multi-label spacecraft electrical signal classification method based on DBN and random forest

• DOI: 10.1371/journal.pone.0176614
• Publication Date: 2017-05-09T15:18:42Z
• Authors: Ke Li; Nan Yu; Pengfei Li; Shimin Song; Yalei Wu; Yang Li; Meng Liu
• Keywords: Deep belief network; SIGNAL (programming language); Feature (linguistics)

In spacecraft electrical signal characteristic data, there exists a large amount of data with high-dimensional features, high computational complexity degree, and low rate identification problems, which causes great difficulty in fault diagnosis electronic load systems. This paper proposes feature extraction method that is based on deep belief networks (DBN) classification the random forest (RF) algorithm; The proposed algorithm mainly employs multi-layer neural network to reduce dimension original then, applied. Firstly, we use wavelet denoising, was used pre-process data. Secondly, improve for characteristics Finally, classify comparing it other algorithms. experimental results show compared algorithms, shows excellent performance terms accuracy, efficiency, stability addressing