In the overload scenario of the LTE-A network

the bursty access of the machine type communication (MTC) device may cause serious congestion or even deadlock in the network

resulting in low network access efficiency. Under the premise that available preambles are limited

controlling the number of devices that initiate the access according to the real-time load can effectively reduce the collision probability of the preamble

but the control method is not clear. To do this

this paper proposes a dynamic access class barring (ACB) mechanism to optimize the random access performance of the massive MTC. An estimation model based on back-off prediction is established.The model estimates the number of real-time active devices based on the number of retransmitted devices and the state transition process. Moreover

combined with the adjustment of the ACB parameter

the number of successfully accessed devices in real time can be optimized

and the access success rate of the device is effectively improved. The performance of the proposed dynamic ACB scheme is compared with that of the existing dynamic ACB schemes under different traffic degrees. Simulation results prove that the access success probability of the proposed dynamic ACB scheme is 100%. Simultaneously

the proposed novel scheme can get lower average access delay comparing with the existing dynamic ACB schemes.