Almost blank subframe (ABS) is a classic and efficient technology to cancel interference in cellular networks and could be applied in future networks such as ultra‐density network. However, the existence of ABS will result in the deterioration of the communication capacity of the macro base station (BS), particularly for applications with large number of edge users. To solve this problem, an improved ABS technology with an interference cancelation scheme is proposed. For the considered scenario, the macro BS and the micro BS work in a cooperative manner, where the signal transmitted by the macro BS interferes with the communication between the micro BS and its edge terminals. At the terminal side, both the nonlinearity of the macro BS transmitter and the propagation model between the edge user and the macro BS are estimated using neural networks (NNs), which are then sent to the micro BS for interference elimination through the ABS. At the micro BS side, firstly, the nonlinearity and channel parameters from the ABS frame are used to generate a cancelation signal. Secondly, the cancelation signal is sent from the micro BS to the edge user to eliminate the disturbance from the macro BS. This enhanced technology can alleviate the disturbance for the edge user. Meanwhile, since it massively decreases the occupancy of ABS in the macro BS, the capacity of the macro cell can be improved.