Authors
Wenhui Zou, Huiyun Zhang, Zhangguang Ni, Meicun Wang, Yan Jiang, Ling Dong, Linyan Zhou
Published in
Food chemistry. Volume 533. Issue Pt 3. Pages 151931. Nov 05, 2026. Epub Nov 05, 2026.
Abstract
This study developed an efficient classification approach for 5 mango cultivars by integrating flavoromics with machine learning (ML) through cross-domain transfer learning. E-nose analysis revealed that, although volatile organic compound (VOC) categories were highly similar across cultivars, their intensities and proportions differed significantly. A total of 76 VOCs were identified by HS-SPME-GC-MS, with terpenes (43.54%-81.75%) and alcohols (5.50%-27.51%) as the dominant groups. To address the limited classification performance of E-nose data caused by multicollinearity, an optimized multi-layer perceptron (MLP) incorporating cross-domain transfer learning was developed, enabling the integration of latent chemical information from GC-MS into E-nose data and improving the mean accuracy from 95.33% to 98.00% accuracy. SHAP and correlation analysis identified top key sensors (S1, S5, S9) and VOC markers ((E)-β-Ocimene, γ-Octanoic lactone and γ-Decalactone). Overall, this study provides a strategy for rapid cultivar identification, supporting variety traceability and quality control in mango industry.
PMID:
43033694
Bibliographic data and abstract were imported from PubMed on 15 Nov 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 6
- Comments 0