Widespread researches and studies on earthquake prediction show that seismoionospheric disturbances can be observed over seismic regions before an earthquake. However, it is still hard to detect accurate pre-seismic ionosphere anomaly and use them to predict earthquake. To solve the problem, we propose a method that can extract the feature of pre-seismic ionospheric anomalies based on data mining. The main theme of this method can be described as follows: First, we mine frequent itemsets from pre-seismic ionosphere data measured by DEMETER (Detection of Electromagnetic Emissions Transmitted from Earthquake Regions) satellite and regard them as seismic features after a series of processing; then we carry out earthquake prediction experiments in test region to check the validity of these features as well as searching for interesting information. The experimental results reveal that this method is effective, and we also get some remarkable conclusions about earthquake prediction, such as the features of pre-seismic ionosphere anomalies contain ion temperature, electron density, electron temperature and plasma potential and among all the experimental conditions we discussed in this paper, and features mined from ionosphere data 5 days before earthquake lead to the highest accuracy in earthquake prediction.

Abstract.Many observational results have shown electromagnetic abnormality in the ionosphere before large earthquakes. The theoretical simulation can help us to understand the internal mechanism of these anomalous electromagnetic signals resulted from seismic regions. In this paper, the horizontal and vertical components of electric and magnetic field at the topside ionosphere are simulated by using the full wave method that is based on an improved transfer matrix method in the lossy anisotropic horizontally stratified ionosphere. Taken account into two earthquakes with electric field perturbations recorded by the DEMETER satellite, the numerical results reveal that the propagation and penetration of ULF (ultra-low-frequency) electromagnetic waves into the ionosphere is related to the spatial distribution of electron and ion densities at different time and locations, in which the ion density has less effect than electron density on the field intensity. Compared with different frequency signals, the minimum values of electric and magnetic field excited by earthquakes can be detected by satellite in current detection capability have also been calculated, and the lower frequency wave can be detected easier.