This paper presents a data-driven optimization-based approach to allocate chargers for battery electric vehicle (BEV) taxis throughout a city with the objective of minimizing the infrastructure investment. To account for charging congestion, an M/M/x/s queueing model is adopted to estimate the probability of BEV taxis being charged at their dwell places. By means of regression and logarithmic transformation, the charger allocation problem is formulated as an integer linear program (ILP), which can be solved efficiently using Gurobi solver. The proposed method is applied using large-scale GPS trajectory data collected from the taxi fleet of Changsha, China. The key findings from the results include the following: (1) the dwell pattern of the taxi fleet determines the siting of charging stations; (2) by providing waiting spots, in addition to charging spots, the utilization of chargers increases and the number of required chargers at each site decreases; and (3) the tradeoff between installing more chargers versus providing more waiting spaces can be quantified by the cost ratio of chargers and parking spots.