CL eVer: A Feature Subset Selection Technique for Multivariate Time Series

Feature subset selection (FSS) is one of the data pre-processing techniques to identify a subset of the original features from a given dataset before performing any data mining tasks. We propose a novel FSS method for Multivariate Time Series (MTS) based on Common Principal Components, termed CLeVer. It utilizes the properties of the principal components to retain the correlation information among original features while traditional FSS techniques, such as Recursive Feature Elimination (RFE), may lose it. In order to evaluate the effectiveness of our selected subset of features, classification is employed as the target data mining task. Our experiments show that CLeVer outperforms RFE and Fisher Criterion by up to a factor of two in terms of classification accuracy, while requiring up to 2 orders of magnitude less processing time.