模式识别(心理学)
支持向量机
人工智能
橄榄油
主成分分析
化学计量学
风味
电子鼻
聚类分析
化学
数学
色谱法
食品科学
计算机科学
作者
Natalie Gerhardt,Sebastian Schwolow,Sascha Rohn,Pilar Ruiz Pérez‐Cacho,Hortensia Galán‐Soldevilla,Lourdes Arce,Philipp Weller
出处
期刊:Food Chemistry
[Elsevier]
日期:2019-04-01
卷期号:278: 720-728
被引量:126
标识
DOI:10.1016/j.foodchem.2018.11.095
摘要
For the first time, this study describes a HS-GC-IMS strategy for analyzing non-targeted volatile organic compounds (VOCs) profiles to distinguish between virgin olive oils of different classification. Correlations among measured flavor characteristics and sensory attributes evaluated by a test panel were determined by applying unsupervised (PCA, HCA) and supervised (LDA, kNN and SVM) chemometric techniques. PCA and HCA were applied for natural clustering of the samples and LDA, kNN, and SVM methods were used to create predictive models for olive oil classification. Identification of 26 target compounds revealed which compounds are responsible for discrimination, and how their distribution correlates with the sensory evaluation. In the investigated samples, LDA, kNN, and SVM models correctly classified 83.3%, 73.8%, and 88.1% of the samples, respectively. This suggests that mathematical correlations of HS-GC-IMS 3D fingerprints with the sensory analysis may be appropriate for calculating a good predictive value to classify virgin olive oils.
科研通智能强力驱动
Strongly Powered by AbleSci AI