A0575
Title: An interpretable sparse estimator for large-scale network autoregressive models
Authors: Simon Trimborn - National University of Singapore (Singapore) [presenting]
Abstract: The era of Big Data requires the discovery of the essential underlying structure. Sparsity estimators imposing structures have the potential to detect (describe) the dynamic dependence while bringing an interpretable system with them. We propose an interpretable sparse estimator which restricts the model for the Lag, Network and individual effects. We derive the theoretical properties and investigate the numerical performance of the estimator in extensive simulations. The sparsity operator facilitates a higher accuracy than alternative ones and simultaneously being fast in computation. We show the applicability of the estimator on a huge network of cryptocurrency pricing series.
