fig1

Figure 1. Overview and architecture of DSMR. (A) The DS module employs a “leave-one-out” approach for gradual elimination and addition, extracting key data samples while preserving the overall data distribution; (B) The MR module integrates algorithm ensemble, automatic parameter tuning, and evaluation ranking, providing a unified modeling and evaluation process to ensure comprehensive and fair assessment of data subsets. DSMR: Data screening and model retrieval framework based on active learning.