We propose a data quality enhancement strategy combining multi-algorithm outlier detection with selective re-experimentation, significantly improving machine learning prediction accuracy for epoxy mechanical properties with minimal additional experimentation.
