Abstract: The granular computing classification algorithms based on distance were proposed. Firstly, granules were represented as the forms of hyperdiamond, hypersphere and hypercube. Secondly, the union operators were designed to realize the transformation between two granule spaces with different granularity. Thirdly, the thresholds of granularity were used to control the union processes. Then the granular computing classification algorithms based on distance were designed. The benchmark datasets in UCI Learning Repository were used to verify the performance of the algorithms, the feasibility of the granular computing classification algorithms was verified by the experimental results