A decomposition method for global optimization, when some of the problem variables are integer, is presented. Potential solutions for a specific problem are encoded into strings that fix the integer variable values. To each string corresponds a continuous sub-problem evaluating the relative quality or fitness of the corresponding string as a solution to the global problem. The search through the solution space defined by the integer variables is conducted heuristically using a genetic algorithm. Computational results are provided for four problems from the literature: a pure 0/1 hub location problem and three mixed integer (0/1) location problems with applications to capacitated plant location, development of offshore oil fields, and skeleton-based facility layout.