MSTL: A Seasonal-Trend Decomposition Algorithm for Time Series with Multiple Seasonal Patterns

The decomposition of time series into components is an important task that helps to understand and can enable better forecasting. Nowadays, with high sampling rates leading high-frequency data (such as daily, hourly, or minutely data), many real-world datasets contain that can exhibit multiple seasonal patterns. Although several methods have been proposed to decompose under these circumstances, they are often computationally inefficient or inaccurate. In this study, we propose Multiple Seasonal-Trend using Loess (MSTL), extension the traditional Seasonal-Trend (STL) procedure, allowing the with our evaluation on synthetic a perturbed dataset, compared other benchmarks, MSTL demonstrates competitive results lower computational cost. implementation available in the R package forecast.