Journal of Chemometrics

Cover image for Vol. 26 Issue 7

July 2012

Volume 26, Issue 7

Pages i–iii, 353–422

  1. Issue Information

    1. Top of page
    2. Issue Information
    3. Research Articles
    1. Issue Information (pages i–iii)

      Version of Record online: 6 JUL 2012 | DOI: 10.1002/cem.2408

  2. Research Articles

    1. Top of page
    2. Issue Information
    3. Research Articles
    1. Application of genetic algorithm-support vector regression (GA-SVR) for quantitative analysis of herbal medicines (pages 353–360)

      Ni Xin, Xiaofeng Gu, Hao Wu, Yuzhu Hu and Zhonglin Yang

      Version of Record online: 27 MAR 2012 | DOI: 10.1002/cem.2435

      Genetic algorithm-support vector regression (GA-SVR) coupled approach is proposed for quantitative analysis of herbal medicines. GA-Random Forests and GA-partial least squares regression were also employed and compared with the GA-SVR method. The performance has been tested on a simulated system, high performance liquid chromatography data set, and near-infrared data set, and the results show that GA-SVR model is helpful for the quantitative analysis of herbal medicines.

    2. Cross-validation in PCA models with the element-wise k-fold (ekf) algorithm: theoretical aspects (pages 361–373)

      José Camacho and Alberto Ferrer

      Version of Record online: 1 MAY 2012 | DOI: 10.1002/cem.2440

      The element-wise k-fold (ekf) cross-validation is among the most used algorithms for principal components analysis cross-validation. It has been stated to outperform other methods under most circumstances in a numerical experiment. A theoretical study is driven to identify in which situations ekf is adequate and in which it is not. The results show that ekf may be unable to assess the extent to which a model represents a test set and may lead to discard principal components with important information.

    3. Active learning for spectroscopic data regression (pages 374–383)

      Fouzi Douak, Farid Melgani, Naif Alajlan, Edoardo Pasolli, Yakoub Bazi and Nabil Benoudjit

      Version of Record online: 19 APR 2012 | DOI: 10.1002/cem.2443

      In this work, we introduce an active learning approach for the estimation of chemical concentrations from spectroscopic data. In particular, we propose two different strategies, developed for regression approaches based on partial least squares regression (PLSR), ridge regression (RR), kernel ridge regression (KRR) and support vector regression (SVR). Experimental results on three different real data sets are reported and discussed.

    4. Quadratic PLS1 regression revisited (pages 384–389)

      Stéphane Verdun, Mohamed Hanafi, Véronique Cariou and El Mostafa Qannari

      Version of Record online: 22 MAY 2012 | DOI: 10.1002/cem.2447

      Within the framework of chemometrics, partial least squares (PLS) has become a standard multivariate tool because it can handle situations dealing with noisy and highly collinear data. However, there are some situations where PLS models do not correctly fit the data because a nonlinearity structure is observed. Within the framework of nonlinear PLS, the quadratic PLS regression approach is discussed. A new algorithm for the determination of the components is presented, and its advantages over the original algorithm are outlined.

    5. Orthogonal signal correction-based prediction of total antioxidant activity using partial least squares regression from chromatograms (pages 390–399)

      Saliha Şahin, Esra Işık, Önder Aybastıer and Cevdet Demir

      Version of Record online: 29 APR 2012 | DOI: 10.1002/cem.2450

      The multivariate calibration methods–PLS, orthogonal signal correction and partial least squares –were employed for the prediction of total antioxidant activities of four Prunella L. species. The importance of the preprocessing was investigated by calculating the root mean square error for the calibration set for the total antioxidant activity. The models developed on the basis of the preprocessed data were able to predict the total antioxidant activity with a precision comparable to that of the reference ABTS and DPPH methods.

    6. A robust regression approach for spectrophotometric signal analysis (pages 400–405)

      Leila Douha, Nabil Benoudjit and Farid Melgani

      Version of Record online: 27 MAY 2012 | DOI: 10.1002/cem.2455

      The effectiveness of a regression method strongly depends on the characteristics of the considered regression problem. As a consequence, this makes it difficult to choose a priori the most appropriate algorithm for a given dataset. This issue is faced in this work through the implementation of the proposed robust multiple system via four different fusion strategies. A novel fusion strategy named selection-based strategy was proposed. The experimental assessment was carried out on the following: a wine, an orange juice, and an apple datasets. The obtained results confirm the superiority of the fusion methods against the traditional one.

    7. Quantifying brain tumor tissue abundance in HR-MAS spectra using non-negative blind source separation techniques (pages 406–415)

      Anca Ramona Croitor Sava, M. Carmen Martinez-Bisbal, Diana Maria Sima, Jorge Calvar, Vicente Esteve, Bernardo Celda, Uwe Himmelreich and Sabine Van Huffel

      Version of Record online: 19 JUN 2012 | DOI: 10.1002/cem.2456

      Blind source separation (BSS) techniques have the potential to identify within high-resolution magic angle spinning (HR-MAS) spectra coming from adult patients presenting a glial tumor the different tumor tissue types that contribute to the profile of each spectrum and to quantify their contribution (abundance) to the profile of each spectrum. The BSS results are in agreement with the pathology obtained by the histopathological examination that succeeded the HR-MAS measurements.

    8. A restricted b-spline basis for S-shaped calibration curves (pages 416–422)

      Geoff Jones and Emily Kawabata

      Version of Record online: 3 JUN 2012 | DOI: 10.1002/cem.2457

      S-shaped curves in calibration assays are commonly modeled using an empirically chosen nonlinear function. We propose an alternative approach using a b-spline basis adapted to produce lower and upper asymptotes in the fitted curves. The model is linear in the parameters and so is numerically stable when fitting mixed models to multiple curves. We explore the performance of this approach using data from a single radioimmunoassay for cortisol and from a set of 96 ELISA assays for the herbicide atrazine.