Feature selection using distributions of orthogonal PLS regression vectors in spectral data

Abstract

Feature selection, which is important for successful analysis of chemometric data, aims to produce parsimonious and predictive models. Partial least squares (PLS) regression is one of the main methods in chemometrics for analyzing multivariate data with input X and response Y by modeling the covariance structure in the X and Y spaces. Recently, orthogonal projections to latent structures (OPLS) has been widely used in processing multivariate data because OPLS improves the interpretability of PLS models by removing systematic variation in the X space not correlated to Y. The purpose of this paper is to present a feature selection method of multivariate data through orthogonal PLS regression (OPLSR), which combines orthogonal signal correction with PLS. The presented method generates empirical distributions of features effects upon Y in OPLSR vectors via permutation tests and examines the significance of the effects of the input features on Y. We show the performance of the proposed method using a simulation study in which a three-layer network structure exists in compared with the false discovery rate method. To demonstrate this method, we apply it to both real-life NIR spectra data and mass spectrometry data.

Machine Learning Self-Diffusion Prediction for Lennard-Jones Fluids in Pores

“…The features were checked for correlations as highly correlated features would not only be redundant and unnecessary, supplying no new information, but could also degrade the performance, reduce the interpretability, and lead to overfitting of the ML model. , Figure a shows a correlation matrix where the Pearson correlation coefficient (eq ) between any two features is indicated. Here, we can see that there is a strong correlation between ρ, D 0, η, and all RDF values, as well as a slightly weaker but significant correlation of T with the aforementioned features.…”

Multivariate Statistical Analysis Uncovers Spectrum–Effect Relationship between HPLC Fingerprints and Antioxidant Activity of Saffron

“…Permutation test, a computer-based resampling method for remodeling and predicting, was widely used in the computation of variable importance and confidence intervals [43,44]. It could be considered that the model had not been overfitted when the Y-axis intercept of R 2 and Q 2 for the established OPLS-DA models was less than 0.3 and 0.05, respectively [45][46][47].…”

Predictive Error Compensating Wavelet Neural Network Model for Multivariable Time Series Prediction

“…Uncovering hidden patterns associated with the target variables is not trivial [10]. One of the major issues in addressing this problem is how to deal with the existence of a high correlation between supplementary and target data, which can cause a problem in overfitting.…”

“…The highest correlating data to the residual error is selected as supplementary data in the second network. Generally, one or two additional feature components are sufficient [10], however, the proposed PEC-WNN model is able to select as many features as supplementary data correlates to the residual error.…”