There are two main mechanisms of miRNA-mediated gene silencing: either mRNA degradation or translational repression. However, the precise mechanism of target mRNAs regulated by miRNA remains unclear. As a complementary approach to experiment, a computational method was proposed to recognize the mechanism of miRNA-mediated gene silencing in human. We have analyzed extensive features correlated with miRNA-mediated silencing mechanism of mRNA. It is found that, the duplex structure, the number of binding sites and the structural accessibility of target site region are effective factors in determining whether a target mRNA is cleaved or only translationally inhibited. An SVM-based classifier was constructed to predict the regulation mode of miRNA based on these informative features. The results indicated that the approach proposed is effective in distinguishing whether a target mRNA is cleaved or translationally inhibited in human. Furthermore, the web server microDoR (http://reprod.njmu.edu.cn/microdor) has been developed and is freely available for users.