An efficient splitting strategy is important for the islanding operation of power systems. In this paper, a mixed-integer linear programming (MILP)-based splitting method is proposed. First, the graph theory is employed to transform the splitting problem into a graph partition problem. Then, the graph partition problem is modeled as an MILP optimization problem with consideration to the network connectivity of all subgraphs. The MILP-based splitting strategy can be efficiently computed with existing commercial solvers, and the multiple optimal solutions can be captured with the recursive process. Compared with the splitting methods in the literature, the proposed method is more flexible and efficient, such that the users can select the optimal solution from multiple solutions with different interests and heuristic splitting rules (e.g., with the minimum amount of switched lines). Finally, the effectiveness of the proposed method has been verified with IEEE 30-, 118-, and 300-bus test systems.