Abstract: Multiple-view visualization recommendation aims to automatically generate and evaluate multi-view de-signs, and recommend suitable solutions to users. To address the limitations of existing methods in handling irregular boundary scenarios and their insufficient consideration of data characteristics, this study propos-es a data-driven framework for the automatic generation and recommendation of multiple-view visualiza-tions tailored to irregular boundaries. First, a data information recommendation module is designed to se-lect high-quality data based on biased entropy-based recommendations. Second, a layout generation mod-ule adapted to irregular boundaries is developed to achieve automated layout construction under complex spatial constraints. Third, a data-guided matching optimization module is introduced to maximize infor-mation utility and enable intelligent alignment between data paths and layouts. Finally, these three modules are integrated into a user-friendly visualization system, where case studies, quantitative experiments, and user evaluations are conducted to validate the framework’s effectiveness. Results show that the proposed framework extends the application scope of traditional multiple-view visualization design and improves both efficiency and quality in irregular boundary scenarios.