Noslen Hernández

Support vector regression for functional data in multivariate calibration problems

Quantitative analyses involving instrumental signals, such as chromatograms, NIR, and MIR spectra have been successfully applied nowadays for the solution of important chemical tasks. Multivariate calibration is very useful for such purposes and the commonly used methods in chemometrics consider each sample spectrum as a sequence of discrete data points. An alternative way to analyze spectral data is to consider each sample as a function, in which a functional data is obtained. Concerning regression, some linear and nonparametric regression methods have been generalized to functional data. This paper proposes the use of the recently introduced method, support vector regression for functional data (FDA-SVR) for the solution of linear and nonlinear multivariate calibration problems. Three different spectral datasets were analyzed and a comparative study was carried out to test its performance with respect to some traditional calibration methods used in chemometrics such as PLS, SVR and LS-SVR. The satisfactory results obtained with FDA-SVR suggest that it can be an effective and promising tool for multivariate calibration tasks.

Support vector regression methods for functional data

Many regression tasks in practice dispose in low gear instance of digitized functions as predictor variables. This has motivated the development of regression methods for functional data. In particular, Naradaya-Watson Kernel (NWK) and Radial Basis Function (RBF) estimators have been recently extended to functional nonparametric regression models. However, these methods do not allow for dimensionality reduction. For this purpose, we introduce Support Vector Regression (SVR) methods for functional data. These are formulated in the framework of approximation in reproducing kernel Hilbert spaces. On this general basis, some of its properties are investigated, emphasizing the construction of nonnegative definite kernels on functional spaces. Furthermore, the performance of SVR for functional variables is shown on a real world benchmark spectrometric data set, as well as comparisons with NWK and RBF methods. Good predictions were obtained by these three approaches, but SVR achieved in addition about 20% reduction of dimensionality.

Dissimilarity representation on functional spectral data for classification

In chemometrics, spectral data are typically represented by vectors of features in spite of the fact that they are usually plotted as functions of e.g. wavelengths and concentrations. In the representation, this functional information is thereby not reflected. Consequently, some characteristics of the data that can be essential for discrimination between samples of different classes or any other analysis are ignored. Examples are the continuity between measured points and the shape of curves. In the Functional Data Analysis (FDA) approach, the functional characteristics of spectra are taken into account by approximating the data by real valued functions, e.g. splines. Another solution is the Dissimilarity Representation (DR), in which classifiers are trained in a space built by dissimilarities with training examples or prototypes of each class. Functional information may be incorporated in the definition of the dissimilarity measure. In this paper we compare the feature‐based representation of chemical spectral data with three other representations: FDA, DR defined on raw data and DR defined on FDA descriptions. We analyze the classification results of these four representations for five data sets of different types, by using different classifiers. We demonstrate the importance of reflecting the functional characteristics of chemical spectral data in their representation, and we show when the presented approaches are more suitable. Copyright

A non-Bayesian predictive approach for statistical calibration

A non-Bayesian predictive approach for statistical calibration is introduced. This is based on particularizing to the calibration setting the general definition of non-Bayesian (or frequentist) predictive probability density proposed by Harris [Predictive fit for natural exponential families, Biometrika 76 (1989), pp. 675–684]. The new method is elaborated in detail in case of Gaussian linear univariate calibration. Through asymptotic analysis and simulation results with moderate sample size, it is shown that the non-Bayesian predictive estimator of the unknown parameter of interest in calibration (commonly, a substance concentration) favourably compares with previous estimators such as the classical and inverse estimators, especially for extrapolation problems. A further advantage of the non-Bayesian predictive approach is that it provides not only point estimates but also a predictive likelihood function that allows the researcher to explore the plausibility of any possible parameter value, which is also briefly illustrated. Furthermore, the introduced approach offers a general framework that can be applied for calibrating on the basis of any parametric statistical model, so making it potentially useful for nonlinear and non-Gaussian calibration problems.

A functional density-based nonparametric approach for statistical calibration

In this paper a new nonparametric functional method is introduced for predicting a scalar random variable Y from a functional random variable X. The resulting prediction has the form of a weighted average of the training data set, where the weights are determined by the conditional probability density of X given Y, which is assumed to be Gaussian. In this way such a conditional probability density is incorporated as a key information into the estimator. Contrary to some previous approaches, no assumption about the dimensionality of E(X|Y = y) is required. The new proposal is computationally simple and easy to implement. Its performance is shown through its application to both simulated and real data.

On Combining Wavelets Expansion and Sparse Linear Models for Regression on Metabolomic Data and Biomarker Selection

Wavelet thresholding of spectra has to be handled with care when the spectra are the predictors of a regression problem. Indeed, a blind thresholding of the signal followed by a regression method often leads to deteriorated predictions. The scope of this article is to show that sparse regression methods, applied in the wavelet domain, perform an automatic thresholding: the most relevant wavelet coefficients are selected to optimize the prediction of a given target of interest. This approach can be seen as a joint thresholding designed for a predictive purpose. The method is illustrated on a real world problem where metabolomic data are linked to poison ingestion. This example proves the usefulness of wavelet expansion and the good behavior of sparse and regularized methods. A comparison study is performed between the two-steps approach (wavelet thresholding and regression) and the one-step approach (selection of wavelet coefficients with a sparse regression). The comparison includes two types of wavelet bases, various thresholding methods, and various regression methods and is evaluated by calculating prediction performances. Information about the location of the most important features on the spectra was also obtained and used to identify the most relevant metabolites involved in the mice poisoning.

Regularized sliced inverse regression for determination of the percentage of crystallinity in FCC catalysts

This paper proposes a new calibration model based on regularized sliced inverse regression (RSIR) for predicting the percentage of crystallinity of fluidized catalytic cracking catalysts (FCC) using Fourier transform mid‐infrared spectroscopy (FT‐MIR). RSIR is an effective dimension‐reduction tool that looks for a proper dimension‐reduction subspace without requiring a pre‐specified functional form for the relation between independent and dependent variables. Combinations of RSIR with linear and nonlinear learning algorithms like multiple linear regression (MLR) and Support vector regression (SVR) were applied to the preprocessed data set. A comparison of performance among the different approaches, including previous results reached using PLS, was done. RSIR–MLR achieved the highest prediction accuracy, leading to a simple calibration model. Copyright