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

摘要

Feature selection is a technique to select a subset of variables which are useful in predicting target responses. From a machine learning viewpoint, irrelevant features in a prediction model deteriorate its generalization ability, and it is critical to remove such redundant features to keep the model from being misled by inappropriate information. The task of feature selection, one of the central tasks in machine learning, helps to reduce overfitting by eliminating redundant features. The prevalence of high-dimensional data becomes a challenge for researchers to perform feature selection [1]. In biological fields, particulary, high dimensionality often arises, and irrelevant and redundant features make up a high proportion of the total data [2].