Statistical and Machine Learning forecasting methods: Concerns and ways forward

Summary

Machine Learning (ML) methods have been proposed in the academic literature as alternatives to statistical ones for time series forecasting. Yet, scant evidence is available about their relative performance in terms of accuracy and computational requirements. The purpose of this paper is to evaluate such performance across multiple forecasting horizons using a large subset of 1045 monthly time series used in the M3 Competition. After comparing the post-sample accuracy of popular ML methods with that of eight traditional statistical ones, we found that the former are dominated across both accuracy measures used and for all forecasting horizons examined. Moreover, we observed that their computational requirements are considerably greater than those of statistical methods. The paper discusses the results, explains why the accuracy of ML models is below that of statistical ones and proposes some possible ways forward. The empirical results found in our research stress the need for objective and unbiased ways to test the performance of forecasting methods that can be achieved through sizable and open competitions allowing meaningful comparisons and definite conclusions.

Related articles

• Every day, millions of footsteps echo across schools, walkwa

  Every day, millions of footsteps echo across schools, walkways, and cities, vanishing into silence — wasted, unnoticed, forgotten; but what if those footsteps carried hidden power, what if ordinary movement could be

• System Development and System Analyst

  This document discusses the processes and roles involved in system development and analysis, explaining the importance of systems analysts in improving business operations through technology.

• Aircraft Reactor Experiment

  This document discusses the Aircraft Reactor Experiment (ARE) conducted in 1954, detailing the construction and operation of a molten-salt-fueled reactor at high temperatures and its components including fuel passages and cooling systems.

• Controlled Components for Internet of Things As-A-Service

  This article introduces an approach using controlled components to assist developers in creating IoT services that incorporate non-functional aspects, enabling service composition and management.

• o9 Solutions Demonstrates Over 10x Faster Optimization for Enterprise Supply Chain Planning with NVIDIA cuOpt

  This document details the announcement from o9 Solutions about the enhanced optimization capabilities of their platform when integrated with NVIDIA cuOpt, showcasing a significant reduction in solve times for large-scale Linear Programming supply chain challenges.