In this paper, we address the problem of interference alignment (IA) over MIMO interference channels with limited channel state information (CSI) feedback based on quantization codebooks. Due to limited feedback and hence imperfect IA, there are residual interferences across different links and different data streams. As a result, the performance of IA is greatly related to the CSI accuracy (namely number of feedback bits) and the number of data streams (namely transmission mode). In order to improve the performance of IA, it makes sense to optimize the system parameters according to the channel conditions. Motivated by this, we first give a quantitative performance analysis for IA under limited feedback, and derive a closed-form expression for the average transmission rate in terms of feedback bits and transmission mode. By maximizing the average transmission rate, we obtain an adaptive feedback allocation scheme, as well as a dynamic mode selection scheme. Furthermore, through asymptotic analysis, we obtain several clear insights on the system performance, and provide some guidelines on the system design. Finally, simulation results validate our theoretical claims, and show that obvious performance gain can be obtained by adjusting feedback bits dynamically or selecting transmission mode adaptively.