According to variation differences of high frequency and low frequency components of coal mine machinery vibration signal, a combined vibration signal prediction method of coal mine machinery based on empirical mode decomposition (EMD) and support vector machine (SVM) is proposed. The vibration signal of rolling bearing is decomposed by EMD to obtain relatively stable instrinsic mode function (IMF) components, and the IMF components with similar degree of the fluctuation are reconstructed to obtain high-frequency and low-frequency subsequences. The high-frequency subsequence and low-frequency subsequence are predicted by SVM respectively, and then the final prediction value is obtained after superposing the two prediction results. The bearing experimental data are selected to verify effectiveness of the method. The results show that the root mean square error, average absolute error and average absolute percentage error of the method are smaller than that of the direct prediction method.The results show that the root mean square error, average absolute error and average absolute percentage error of the combined predition method are all smaller than those of direct prediction method. The combined prediction method is applied to condition prediction of rolling bearing of the belt conveyor in main shaft of a coal preparation plant, and the prediction results are consistent with actual situation.