Selecting the Top-k Discriminative Features Using Principal Component Analysis

Feature selection is important for dimensionality reduction, analysis, and pattern discovery applications. We consider multivariate time series data and propose an unsupervised learning technique to identify the top-k discriminative features. The proposed technique uses statistics drawn from the Principal Component Analysis (PCA) of the input data to leverage the relative importance of the principal components along with the coefficients within the principal directions of the data to uncover the ranking of the features. We conduct numerous experiments using various benchmark datasets to study the performance of the proposed technique in terms of the discriminant power of the selected features and its ability to minimize the original data reconstruction error. Compared to major existing techniques, our results indicate increased accuracy and efficiency. We also show that our technique yields improved classification accuracy.