Abstract: The main factors inducing reservoir earthquakes, including lithology, rock mass integrity, fault property, tectonic stress state and seismic activity background in reservoir area, are divided into 11 factors and are quantified. Fuzzy factor of each sample reflecting the effect of the sample on this hyperplane was calculated based on the distance to the hyperplane of each class samples. The Fuzzy Support Vector Machines (FSVM) and Support Vector Machines (SVM) are used to establish the classifier of the induced earthquake, and to predict the magnitude of the reservoir induced earthquake (RIE). The cases analysis shows that FSVM and SVM models can be employed to predict the magnitude of RIE with high precision and over-all consideration. The SVM model are slightly superior to FSVM in the field of RIE prediction. Furthermore, when SVM and FSVM model are applied to classify samples, the SVM model is superior to FSVM if samples are with high discreteness. On the contrary, the FSVM model is superior to SVM if samples are with low discreteness.