Time Series Analysis: yesterday, today, tomorrow

TL;DR

This paper evaluates the effectiveness of modern deep learning and kernel methods in time series forecasting compared to traditional statistical approaches, questioning the assumed superiority of newer techniques.

Contribution

It provides an analysis comparing classical statistical methods with recent deep learning and kernel approaches in time series forecasting.

Findings

Deep learning and kernel methods show improved accuracy in some cases.

Classical methods remain competitive in certain scenarios.

The paper challenges the assumption that newer methods are always superior.

Abstract

Forecasts of various processes have always been a sophisticated problem for statistics and data science. Over the past decades the solution procedures were updated by deep learning and kernel methods. According to many specialists, these approaches are much more precise, stable, and suitable compared to the classical statistical linear time series methods. Here we investigate how true this point of view is.