Two-Step Meta-Learning for Time-Series Forecasting Ensemble
This research provides a more robust and adaptive time-series forecasting solution for businesses and researchers dealing with diverse time-series data, offering improved accuracy over existing benchmarks.
This paper addresses the challenge of time-series forecasting by proposing a two-step meta-learning approach to adaptively determine ensemble diversity and size. The method, tested on 38633 micro-economic time-series from the M4 competition, outperformed Theta and Comb benchmarks, achieving a symmetric mean absolute percentage error of 9.21% with weighted pooling, compared to 11.05% for the Theta method.
Amounts of historical data collected increase and business intelligence applicability with automatic forecasting of time series are in high demand. While no single time series modeling method is universal to all types of dynamics, forecasting using an ensemble of several methods is often seen as a compromise. Instead of fixing ensemble diversity and size, we propose to predict these aspects adaptively using meta-learning. Meta-learning here considers two separate random forest regression models, built on 390 time-series features, to rank 22 univariate forecasting methods and recommend ensemble size. The forecasting ensemble is consequently formed from methods ranked as the best, and forecasts are pooled using either simple or weighted average (with a weight corresponding to reciprocal rank). The proposed approach was tested on 12561 micro-economic time-series (expanded to 38633 for various forecasting horizons) of M4 competition where meta-learning outperformed Theta and Comb benchmarks by relative forecasting errors for all data types and horizons. Best overall results were achieved by weighted pooling with a symmetric mean absolute percentage error of 9.21% versus 11.05% obtained using the Theta method.