期刊
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
卷 118, 期 -, 页码 301-307出版社
ELSEVIER SCIENCE BV
DOI: 10.1016/j.chemolab.2012.05.008
关键词
PLS; LASSO; Proton transfer reaction-mass spectrometry; Time-of-flight; Prediction; Multivariate correlation
类别
资金
- Autonomous Province of Trento, Italy
Recently the first applications in food science and technology of the newly available volatile organic compound (VOC) detection technique proton transfer reaction-mass spectrometry, coupled with a time of flight mass analyzer (PTR-TOF-MS), have been published. In comparison with standard techniques such as GC-MS. PTR-TOF-MS has the remarkable advantage of being extremely fast but has the drawback that compound identification is more challenging and often not possible without further information. In order to better exploit and understand the analytical information entangled in the PTR-TOF-MS fingerprint and to link it with SPME/GC-MS analyses we employed two multivariate calibration methods, PLS and the more recent LASSO. We show that, while in some cases it is sufficient to consider a single PTR-TOF-MS peak in order to predict the intensity of a SPME/GC-MS peak, in general a multivariate approach is needed. We compare the performances of PLS and LASSO in terms of prediction capabilities and interpretability of the model coefficients and conclude that LASSO is more suitable for this problem. As case study, we compared GC and PTR-MS data for different matrices, namely olive oil and grana cheese. (C) 2012 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据