Natural or social disasters can collapse the traditional cellular communications network. To provide continuous communication services as well as support for rescue operations, an unmanned aerial vehicle (UAV) is utilized to replace the traditional base stations. In this paper, we propose a joint UAV deployment and transmit power optimization to maximize the number of users served during the disasters. To solve the optimization problem, we introduce a two-phase algorithm in which the first phase implements the mini batch K-means method. In the second phase, the linear search is used to find the optimal transmit power and the height of UAVs. Numerical results reveal that our proposed approach using the mini batch K-means is outperformed the ones using the traditional K-means and K-medoids methods in terms of clustering time, the number of served users, and the serving time.