Anomaly prediction in seismic signals using neural networks

In this paper, we present a robust technique for predicting anomalies in the near future of an observed signal. First, wavelet de-noising is applied to the signal. Next, peak-finding algorithms search for smaller anomalies that appear frequently throughout the signal. Then the data from the peak-finding algorithm is fed into a feed-forward neural which predicts the likelihood of an anomalous event occurring later in the signal. The neural network is trained using supervised learning techniques with data sets consisting of a mix of signals known to precede anomalous events, and signals known to be free of significant anomalies. Our approach provides a means of predicting large events in signals such as seismograms, EKGs, EEGs, and other non-stationary signals. The proposed technique yielded 83% accuracy when used to predict earthquakes using seismic signals, and so is an effective strategy for predicting seismic events.