The next generation 3GPP long-term evolution (LTE) mobile system will use orthogonal frequency division multiplexing access (OFDMA) as the key technology to improve the spectrum efficiency and flexible user resource allocation. One of the challenges of OFDMA systems is the interference at the cell edge where users are likely to get high interference from neighbouring cells. In a network with non-uniform traffic distribution, the problem becomes even more challenging. In this paper, a semi-smart antenna based resource allocation scheme is proposed for the LTE OFDMA system. The scheme features a centralised genetic algorithm (GA) algorithm that changes the cellular coverage in a coordinated manner to improve the system capacity and mitigate network interference. Simulation experiments have been carried out to evaluate the performance of the proposed scheme, the results showing significant performance improvement in terms of total traffic load achieved for all cells compared to a system with fixed coverage pattern.