Forecasting with Multiple Seasonality

TL;DR

This paper introduces a two-stage forecasting method for time series with multiple seasonality, which does not require prior knowledge of seasonality periods and outperforms existing models like Facebook Prophet.

Contribution

It generalizes the classical ARMA model to handle multiple seasonality without pre-determined periods and employs a new lag order selection criterion.

Findings

Demonstrates superior predictive accuracy over Facebook Prophet.

Effective in capturing complex multiple seasonality patterns.

Applicable to both simulated and real-world data.

Abstract

An emerging number of modern applications involve forecasting time series data that exhibit both short-time dynamics and long-time seasonality. Specifically, time series with multiple seasonality is a difficult task with comparatively fewer discussions. In this paper, we propose a two-stage method for time series with multiple seasonality, which does not require pre-determined seasonality periods. In the first stage, we generalize the classical seasonal autoregressive moving average (ARMA) model in multiple seasonality regime. In the second stage, we utilize an appropriate criterion for lag order selection. Simulation and empirical studies show the excellent predictive performance of our method, especially compared to a recently popular `Facebook Prophet' model for time series.