Carbon dots (CDs) derived from biomass waste represent a sustainable alternative to conventional fluorescent nanomaterials. In this study, highly fluorescent CDs were synthesized from banana peel by pyrolysis using an integrated approach combining Central Composite Design (CCD) and machine learning (ML). The workflow included precursor screening, CCD-based optimization, ML prediction of fluorescence intensity, and environmental stability evaluation. Among the tested precursors, urea was the most effective for enhancing fluorescence performance and quantum yield. CCD identified the key synthesis parameters affecting fluorescence, and the optimized conditions yielded a predicted fluorescence intensity of 16, 696 a. u. , with quantum yield approaching 40%. Among seven ML algorithms, Random Forest showed the best overall predictive performance, achieving the highest cross-validated R 2 of 0. 97 with a low RMSE of 542. 14. Feature importance analysis revealed that measured carbon content (C0) was the dominant predictor when elemental descriptors were included, while the precursor-to-banana-peel mass ratio (mP2) was the most important controllable synthesis factor. Urea-derived BP-CDs retained approximately 97% fluorescence intensity at 1. 0 M NaCl and maintained stable emission across pH 3–11. These findings demonstrate a data-guided strategy for producing high-performance, environmentally robust CDs from agricultural waste.
Nguyen et al. (Fri,) studied this question.