Two-stage models for forecasting time series with multiple seasonality

Abstract

Complex multiple seasonality is an important emerging challenge in time series forecasting. In this paper, we propose models under a framework to forecast such time series. The framework segregates the task into two stages. In the firrst stage, the time series is aggregated and existing time series models such as regression, Box-Jenkins or TBATS, are used to fit this lower frequency data. In the second stage, additive or multiplicative seasonality at the higher frequency levels may be estimated using classical, or function-based methods. Finally, the estimates from the two stages are combined. Detailed illustration is provided via energy load data in New York, collected at five minute intervals. The results are encouraging in terms of computational speed and forecast accuracy as compared to available alternatives.