Feature Selection and Identification of Underground Nuclear Explosion and Natural Earthquake Based on Gamma Test and BP Neural Network

Abstract

Feature selection is a very important and difficult problem in the identification of underground nuclear explosions and natural earthquakes. To solve this, Gamma test is proposed to select a best feature set from all features of underground nuclear explosions and natural earthquakes in the sense of the smallest estimated mean-squared error between feature input and target output, and then an identification experiment based on BP Neural Network is carried on with these selected features. To show the advantages of this method, all features are also identified based on BP Neural Network, the result is that these two identification rates are almost the same, this fully indicates this feature selection and identification method can reduce the complexity of identification system, and improve the efficiency of classification.