Discriminating earthquakes and explosion events by seismic signals basing on BP-Adaboost classifier

Aiming to discriminate earthquakes and explosion events by seismic signals, this paper makes comparisons among Back-propagation neural networks(BP-NN), support vector machine(SVM), and an ensemble learning method: BP-Adaboost. Experiments on our wave features dataset extracted from seismic signals of earthquake and explosion events, have shown that BP-Adaboost can achieve the overall correct recognition rate not less than 98%, with excellent generalization capability, obviously surpassing BP-NN and SVM. The explanation of BP-Adaboost superiors to other two classical main methods is also briefly given.