Traditional negative Poisson’s ratio (NPR) metamaterials consist of identical periodic NPR microstructures, thereby often demonstrating deterministic deformational modes under the given loads. In this paper, a novel data-driven multiscale topology optimization method is developed to design aperiodic NPR metamaterials, enabling flexible control over structural deformational modes. Initially, this method employs the homogenization method to design prototype microstructure with NPR property. Subsequently, to obtain graded microstructures with versatile NPR properties while guaranteeing their connectivity, a progressive optimization strategy is proposed. Furthermore, to enable adaptable adjustments of metamaterials’ mechanical behavior, several heuristically designed graded microstructures are also generated as candidates. Utilizing the property data of these candidate microstructures, a multiscale topology optimization model is constructed to facilitate the aperiodic distribution of NPR microstructures within the NPR metamaterials. Numerical examples are presented to illustrate the effectiveness of the proposed method, showcasing its ability to design NPR metamaterials with tailorable deformational behaviors.