Feature Selection for Multivariate Time Series
Description
Implements feature selection routines for multivariate time series (MTS).
The list of implemented algorithms includes:
own lags (independent MTS components),
distance-based (using external structure, e.g. Pfeifer and Deutsch (1980) ),
cross-correlation (see Schelter et al. (2006, ISBN:9783527406234)),
graphical LASSO (see Haworth and Cheng (2014) ),
random forest (see Pavlyuk (2020) "Random Forest Variable Selection for Sparse Vector Autoregressive Models" in Contributions to Statistics, in production),
least angle regression (see Gelper and Croux (2008) ),
mutual information (see Schelter et al. (2006, ISBN:9783527406234), Liu et al. (2016) ),
and partial spectral coherence (see Davis et al.(2016) ).
In addition, the package implements functions for ensemble feature selection (using feature ranking and majority voting).
The package is implemented within Dmitry Pavlyuk's research project No. 1.1.1.2/VIAA/1/16/112 "Spatiotemporal urban traffic modelling using big data".