Computational cost is a big challenge for almost all intelligent algorithms which are run on CPU. In this regard, our proposed kernel P system multi-objective binary particle swarm optimization feature selection and classification method should perform with an efficient time that we aimed to settle via using potentials of membrane computing in parallel processing and nondeterminism. Moreover, GPUs perform better with latency-tolerant, highly parallel and independent tasks. In this study, to meet all the potentials of a membrane-inspired model particularly parallelism and to improve the time cost, feature selection method implemented on GPU. The time cost of the proposed method on CPU, GPU and Multicore indicates a significant improvement via implementing method on GPU.