Inverse kinematics equations were established for a 3-P-(2U-2S) parallel mechanism of machine tool, which solved for its workspace. By calculating the Jacobian matrix condition number, the condition index of kinematics properties of the parallel mechanism was obtained. By taking the workspace and Global Condition Index (GCI) as the evaluation index, with using of BP Neural network algorithm and genetic algorithm (GA), structure size parameters of radius Rand rod length L (R/L) of the machine tool mechanism were comprehensively optimized. The results show that the performance of the parallel mechanism has been improved after optimization.