Abstract: The principle of support vector machine (SVM) and its regression algorithm is introduced in this paper. The multidimensional samples from theoretical formula are tested by using of SVM. The best model for synthetic prediction of seismic precursors is established according to the SVM algorithm and the seismic precursory anomalies, and it is tested by using of the testing samples. It indicates that the forecast result of the best mode and earthquake magnitude of real seismic examples are basically consistent. It shows that the support vector machine algorithm has an obvious superiority whatever on machine learning or prediction accuracy, and the model for synthetic prediction of seismic precursors based on the SVM theory is feasible, and it can forecast the magnitude of main earthquakes more accurately.

Key words: Support vector machine (SVM), Seismic cursor, Typical earthquake example, Synthetic forecast