Fatoumata Dama

Fatoumata Dama, Christine Sinoquet

Time series modeling for predictive purpose has been an active research area of machine learning for many years. However, no sufficiently comprehensive and meanwhile substantive survey was offered so far. This survey strives to meet this need. A unified presentation has been adopted for entire parts of this compilation. A red thread guides the reader from time series preprocessing to forecasting. Time series decomposition is a major preprocessing task, to separate nonstationary effects (the deterministic components) from the remaining stochastic constituent, assumed to be stationary. The deterministic components are predictable and contribute to the prediction through estimations or extrapolation. Fitting the most appropriate model to the remaining stochastic component aims at capturing the relationship between past and future values, to allow prediction. We cover a sufficiently broad spectrum of models while nonetheless offering substantial methodological developments. We describe three major linear parametric models, together with two nonlinear extensions, and present five categories of nonlinear parametric models. Beyond conventional statistical models, we highlight six categories of deep neural networks appropriate for time series forecasting in nonlinear framework. Finally, we enlighten new avenues of research for time series modeling and forecasting. We also report software made publicly available for the models presented.

Fatoumata Dama, Christine Sinoquet

Time series subject to change in regime have attracted much interest in domains such as econometry, finance or meteorology. For discrete-valued regimes, some models such as the popular Hidden Markov Chain (HMC) describe time series whose state process is unknown at all time-steps. Sometimes, time series are firstly labelled thanks to some annotation function. Thus, another category of models handles the case with regimes observed at all time-steps. We present a novel model which addresses the intermediate case: (i) state processes associated to such time series are modelled by Partially Hidden Markov Chains (PHMCs); (ii) a linear autoregressive (LAR) model drives the dynamics of the time series, within each regime. We describe a variant of the expection maximization (EM) algorithm devoted to PHMC-LAR model learning. We propose a hidden state inference procedure and a forecasting function that take into account the observed states when existing. We assess inference and prediction performances, and analyze EM convergence times for the new model, using simulated data. We show the benefits of using partially observed states to decrease EM convergence times. A fully labelled scheme with unreliable labels also speeds up EM. This offers promising prospects to enhance PHMC-LAR model selection. We also point out the robustness of PHMC-LAR to labelling errors in inference task, when large training datasets and moderate labelling error rates are considered. Finally, we highlight the remarkable robustness to error labelling in the prediction task, over the whole range of error rates.