Geographical origin identification of sweet cherry based on quality traits combined with DD-SIMCA and XGBoost.

Geographical origin identification technologies based on physical and nutritional characteristics have recently been developed and applied. This study evaluated the feasibility of identifying the geographical origin of sweet cherries using organoleptic traits and phenolic compound profiles. Data-driven soft independent modeling of class analogy (DD-SIMCA) and extreme gradient boosting (XGBoost) were applied to 170 sweet cherry samples collected in 2023 and 2024 from Beijing, Dalian, Tianshui, and Yantai, China. Measurements included transverse diameter, longitudinal diameter, fruit weight, soluble solid content, titratable acidity, organic acids, ascorbic acid, and 14 phenolic compounds. The DD-SIMCA model showed high sensitivity (98.00 %) and specificity (100.00 %). XGBoost yielded a prediction accuracy of 94.12 %, outperforming LDA (82.35 %), RF (88.24 %), and k-NN (82.35 %). Key discriminatory features included malic acid, quinic acid, citric acid, kaempferol-3-O-rutinoside, titratable acidity, and cyanidin-3-O-rutinoside. These findings indicate that DD-SIMCA and XGBoost are effective methods for the geographical origin identification of sweet cherries based on quality attributes. This approach supports quality assurance and control in regional production systems.