Anomalous environmental electromagnetic (EM) radiation waves have been reported as the portents of earthquakes. We have been measuring the Extremely Low Frequency (ELF) range all over Japan. Our goal is to predict earthquakes using EM radiation waves. Previously, we proposed a method of detecting anomalous signals by focusing on linear prediction errors. However, this method also sensitively responds to earthquake-unrelated anomaly. For accurate earthquake-prediction, we should eliminate earthquake-unrelated signals. In this paper, we try to reduce false detection rate by global signal elimination using Non-negative Matrix Factorization (NMF) and evaluate the effectiveness of this method.