Machine Learning
ICONIP'06 Proceedings of the 13th international conference on Neural Information Processing - Volume Part II
Orthogonal least squares learning algorithm for radial basis function networks
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
Soft computing techniques are known in scientific literature as capable methods for function approximation. Within this framework, they are applied to forecasting time series in non-linear problems, where estimation of the sample starting from actual measurements is very difficult. In this paper, we exploited soft computing techniques in order to predict the number of earthquakes (i.e. aftershocks) occuring after a large earthquake. The forecasting involves the aftershocks occuring day by day after a large earthquake, i.e. an earthquake having a magnitude M=7.0 Richter. In particular, a comparison between radial basis function neural networks and support vector regression machines has been carried out, in order to overcome some problems related to the so called Delta/Sigma method, i.e. a probabilistic approach already used to detect aftershock events with magnitude M5.5 after a large earthquake. A database for the Pacific area is used for the study, and the obtained results are very interesting.