To calculate the process similarity with consideration of deep semantics correlation between business processes, and to optimize the time complexity and matching result when the node number of business process becomes larger and larger, a process matching method based on GA (Genetic Algorithm) is put forward. This method is applied in similarity calculation for both process semantic and process structure, in which encoding is determined, and greedy algorithm is utilized to initialize the population of GA. By defining genetic operations and adopting some strategies for simplifying, the optimization of business process matching with large node number is fulfilled. As is expected, the experiments prove that the overall performance of algorithm proposed in this paper is better than the others that exist, especially when the count of process nodes grows to a large number. So it is concluded that the application of GA in business process similarity calculation and corresponding process optimization can effectively control the time complexity, meanwhile ensure the quality of the matching result, which shows a good practicability.