To further study the reliability problems of capacitive RF-MEMS switches in practical applications, a multi-core relevance vector machine ( MKRVM) method based on differential evolution quantum particle swarm optimization (DE-QPSO) is proposed to predict the switch lifetime. First of all, bandwidth restricted empirical mode decomposition (BREMD) is used to denoise the life data obtained during the experiment to improve the data reliability; secondly, DE-QPSO is used to obtain the optimal sparse weight of MKRVM, and the MKRVM algorithm is used to predict the life of such switches; finally, the actual data obtained by experiment is used to test the accuracy of the methods. The experimental results show that MKRVM can obtain the prediction results within 0. 21 s. The root mean square of the data is 3. 104 3×10 6 s, which is the closest to the original data of 3. 065 7×10 6 s; DE-QPSO can be optimized within 0. 45 s, the variance is 7×10 -5 . At the same time, it is concluded that the switch life is the longest when the elastic coefficient is in the range of 4~ 16 N/ m.