期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2017
卷号：95
期号：19
页码：5194
出版社：Journal of Theoretical and Applied
摘要：In this paper, Auto-Encoder algorithm (AE) has been used in unsupervised feature selection, then, Back-propagation (BP) algorithm has been used to train reconstructed subsets in supervised learning; in order to recognize human activities inside smart home. Subsequently, the performances of auto-encoder have been evaluated and compared with traditional weighting technique for features selection. The experimental results demonstrate that neural network using auto-encoder achieves an average of over 91.46 % for one user and 90.62 % for two-users, relatively better than neural network using traditional weighting technique.
关键词：Auto-Encoder Pre-Training; Deep Network; Activity Recognition; Back-Propagation Algorithm; Smart Home.
