Aiming at shortcomings of existing mass exchanger network optimization methods, a random walk algorithm with compulsive evolution for mass exchanger network synthesis is proposed, in which the mass transfer load, split ratio and the flow of lean streams of mass exchanger are increased or reduced randomly. A minimum threshold is set to realize synchronous optimization of network continuity and integer variables. A small probability is retained to accept the difference solution. It enhances mutation ability of the structure, and makes the algorithm taking into account global search and local search of mass exchanger network. Application in two mass exchanger network examples show that the optimization results are better than those in current literatures. The algorithm maintains independent evolution between individuals and has good global and local search ability.